starlingx
/
metal
Code Issues Proposed changes
metal/mtce/src/rmon/rmonHdlr.cpp

4919 lines
190 KiB
C++
Raw Blame History

/*
* Copyright (c) 2013-2017 Wind River Systems, Inc.
*
* SPDX-License-Identifier: Apache-2.0
*
*/
/**
* @file
* Wind River CGCS Platform Resource Monitor Handler
*/
#include "rmon.h" /* rmon header file */
#include "rmonHttp.h" /* for rmon HTTP libEvent utilties */
#include "rmonApi.h" /* vswitch calls */
#include <sys/wait.h>
#include <time.h>
#include <signal.h>
#include <fstream>
#include <sstream>
#include <ctime>
#include <vector> /* for storing dynamic resource names */
#include <dirent.h>
#include <algorithm>
#include <fcntl.h>
#include <sys/mman.h>
#include <sys/syscall.h>
#include <cctype>
#include <pthread.h>
#include <linux/rtnetlink.h> /* for ... RTMGRP_LINK */
#include "nlEvent.h" /* for ... open_netlink_socket */
#include "nodeEvent.h" /* for inotify */
#include <json-c/json.h> /* for ... json-c json string parsing */
#include "jsonUtil.h"
#include "tokenUtil.h" /* for ... tokenUtil_new_token */
/* Preserve a local copy of a pointer to the control struct to
* avoid having to publish a get utility prototype into rmon.h */
static rmon_ctrl_type * _rmon_ctrl_ptr = NULL ;
static interface_resource_config_type interface_resource_config[MAX_RESOURCES] ;
static resource_config_type resource_config[MAX_RESOURCES] ;
static thinmeta_resource_config_type thinmeta_resource_config[MAX_RESOURCES] ;
static registered_clients registered_clt[MAX_CLIENTS];
static libEvent_type ceilometerEvent; // for ceilometer REST API request
static libEvent tokenEvent; // for token request
/* Used to set alarms through the FM API */
static SFmAlarmDataT alarmData;
static struct mtc_timer rmonTimer_event ;
static struct mtc_timer rmonTimer_pm ;
static struct mtc_timer rmonTimer_ntp ;
static struct mtc_timer rtimer[MAX_RESOURCES] ;
static struct mtc_timer thinmetatimer[MAX_RESOURCES] ;
static ntpStage_enum ntp_stage ; /* The stage the ntp is in within the resource handler fsm */
static int ntp_status ; /* status returned by the ntpq command */
static int ntp_child_pid ;
/* for dynamic resources */
bool modifyingResources = false;
vector<string> criticality_resource;
vector<string> dynamic_resource;
vector<string> types;
vector<string> devices;
vector<int> fs_index;
vector<string> fs_state;
/** List of config files */
std::list<string> config_files ;
std::list<string>::iterator string_iter_ptr ;
std::list<string> interface_config_files ;
/* percent or abs value for fs resources */
int fs_percent = 0;
int swact_count = 0;
/* for cpu usage */
time_t t1, t2;
int num_cpus = 0;
int num_base_cpus = 0;
int included_cpu[MAX_BASE_CPU];
static string hostUUID = "";
/* Initial cpu time */
vector<unsigned long long> cpu_time_initial;
/* Later cpu time */
vector<unsigned long long> cpu_time_later;
void save_fs_resource ( string resource_name, string criticality,
int enabled, int percent, int abs_values[3],
int alarm_type, string type, string device, int mounted );
void calculate_fs_usage( resource_config_type * ptr );
void _space_to_underscore (string & str );
struct thread_data
{
pid_t tid;
pid_t pid;
unsigned long long nr_switches_count;
bool thread_running;
double resource_usage;
resource_config_type * resource;
};
/* info passed to pthreads */
struct thread_data t_data;
pthread_t thread;
pthread_mutex_t lock;
/* strict memory accounting off = 0 or on = 1 */
int IS_STRICT = 0;
void mem_log_ctrl ( rmon_ctrl_type * ptr )
{
#define MAX_LEN 500
char str[MAX_LEN] ;
snprintf (&str[0], MAX_LEN, "%s %s %s\n",
&ptr->my_hostname[0],
ptr->my_address.c_str(),
ptr->my_macaddr.c_str() );
mem_log(str);
}
void mem_log_resource ( resource_config_type * ptr )
{
#define MAX_LEN 500
char str[MAX_LEN] ;
snprintf (&str[0], MAX_LEN, "Resource:%-15s Sev:%-8s Tries:%u Debounce:%d\n",
ptr->resource, ptr->severity, ptr->count, ptr->debounce);
mem_log(str);
}
void mem_log_interface_resource ( interface_resource_config_type * ptr )
{
#define MAX_LEN 500
char str[MAX_LEN] ;
snprintf (&str[0], MAX_LEN, "Resource:%-15s Sev:%-8s Debounce:%d\n",
ptr->resource, ptr->severity, ptr->debounce);
mem_log(str);
}
int _config_dir_load (void);
int _config_files_load (void);
const char rmonStages_str [RMON_STAGE__STAGES][32] =
{
"Handler-Init",
"Handler-Start",
"Manage-Restart",
"Monitor-Wait",
"Monitor-Resource",
"Restart-Wait",
"Ignore-Resource",
"Handler-Finish",
"Failed-Resource",
"Failed-Resource-clr",
} ;
const char ntpStages_str [NTP_STAGE__STAGES][32] =
{
"Begin",
"Execute-NTPQ",
"Execute-NTPQ-Wait",
} ;
registered_clients * get_registered_clients_ptr ( int index )
{
if ( index <= _rmon_ctrl_ptr->clients )
return ( &registered_clt[index] );
return ( NULL );
}
rmon_ctrl_type * get_rmon_ctrl_ptr ()
{
return _rmon_ctrl_ptr;
}
interface_resource_config_type * get_interface_ptr ( int index )
{
if ( index <= _rmon_ctrl_ptr->interface_resources )
return ( &interface_resource_config[index] );
return ( NULL );
}
resource_config_type * get_resource_ptr ( int index )
{
if ( index >= 0 && index <= _rmon_ctrl_ptr->resources )
return ( &resource_config[index] );
return NULL;
}
/*****************************************************************************
*
* Name : get_resource_index
*
* Purpose : Get the resource's index based on the name
*
*****************************************************************************/
int get_resource_index ( const char *resource_name, int *index )
{
for ( int i = 0 ; i < _rmon_ctrl_ptr->resources ; i++ )
{
if ( strcmp(resource_config[i].resource, resource_name) == 0)
{
*index = i;
return (PASS);
}
}
return (FAIL);
}
/*****************************************************************************
*
* Name : rmon_hdlr_fini
*
* Purpose : Clean up the resource monitor module
*
*****************************************************************************/
void rmon_hdlr_fini ( rmon_ctrl_type * ctrl_ptr )
{
for ( int i = 0 ; i < ctrl_ptr->resources ; i++ )
{
// mem_log ('\n');
mem_log_resource ( &resource_config[i] );
}
pthread_mutex_destroy(&lock);
/* Turn off inotify */
//set_inotify_close ( ctrl_ptr->fd, ctrl_ptr->wd );
}
/*****************************************************************************
*
* Name : resourceStageChange
*
* Purpose : Put a resource in the requested stage for use by the resource handler
*
*****************************************************************************/
int resourceStageChange ( resource_config_type * ptr , rmonStage_enum newStage )
{
if (( newStage < RMON_STAGE__STAGES ) &&
( ptr->stage < RMON_STAGE__STAGES ))
{
clog ("%s %s -> %s (%d->%d)\n",
ptr->resource,
rmonStages_str[ptr->stage],
rmonStages_str[newStage],
ptr->stage, newStage);
ptr->stage = newStage ;
return (PASS);
}
else
{
slog ("%s Invalid Stage (now:%d new:%d)\n",
ptr->resource, ptr->stage, newStage );
ptr->stage = RMON_STAGE__FINISH ;
return (FAIL);
}
}
/*****************************************************************************
*
* Name : ntpStageChange
*
* Purpose : Stage change handler for NTP resource
*
*****************************************************************************/
int ntpStageChange ( ntpStage_enum newStage )
{
if ((newStage < NTP_STAGE__STAGES ) &&
( ntp_stage < NTP_STAGE__STAGES ))
{
clog ("NTP %s -> %s (%d->%d)\n",
ntpStages_str[ntp_stage],
ntpStages_str[newStage],
ntp_stage, newStage);
ntp_stage = newStage ;
return (PASS);
}
else
{
slog ("NTP Invalid Stage (now:%d new:%d)\n", ntp_stage, newStage );
ntp_stage = NTP_STAGE__BEGIN ;
return (FAIL);
}
}
/*****************************************************************************
*
* Name : _config_files_load
*
* Purpose : Load the content of each config file into resource_config[x]
*
*****************************************************************************/
int _config_files_load (void)
{
int i = 0 ;
/* Run Maintenance on Inventory */
for ( string_iter_ptr = config_files.begin () ;
string_iter_ptr != config_files.end () ;
string_iter_ptr++ )
{
if ( i >= MAX_RESOURCES )
{
wlog ("Cannot Monitor more than %d resources\n", MAX_RESOURCES );
break ;
}
/* Read the resource config file */
resource_config[i].mask = 0 ;
if (ini_parse( string_iter_ptr->data(), rmon_resource_config,
&resource_config[i]) < 0)
{
ilog("Read Failure : %s\n", string_iter_ptr->data() );
}
else
{
dlog ("Config File : %s\n", string_iter_ptr->c_str());
/* Init the timer for this resource */
mtcTimer_reset ( rtimer[i] ) ;
rtimer[i].service = resource_config[i].resource ;
resource_config[i].i = i ;
/* allow to clear an existing alarm if the first reading is good
after reboot
*/
resource_config[i].failed = false ;
resource_config[i].count = 0 ;
resource_config[i].resource_value = 0 ;
resource_config[i].resource_prev = 0 ;
resource_config[i].stage = RMON_STAGE__INIT ;
resource_config[i].sev = SEVERITY_CLEARED ;
resource_config[i].alarm_type = STANDARD_ALARM;
resource_config[i].failed_send = 0;
resource_config[i].alarm_raised = false;
/* add the alarm ids for the FM API per resource monitored */
if (strcmp(resource_config[i].resource, CPU_RESOURCE_NAME) == 0) {
/* platform cpu utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, CPU_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__CPU_USAGE ;
}
else if (strcmp(resource_config[i].resource, V_CPU_RESOURCE_NAME) == 0) {
/* vswitch cpu utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, V_CPU_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__CPU_USAGE ;
}
else if (strcmp(resource_config[i].resource, MEMORY_RESOURCE_NAME) == 0) {
/* platform memory utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, MEMORY_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__MEMORY_USAGE ;
}
else if (strcmp(resource_config[i].resource, V_MEMORY_RESOURCE_NAME) == 0) {
/* vswitch memory utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, V_MEMORY_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__MEMORY_USAGE ;
}
else if (strcmp(resource_config[i].resource, FS_RESOURCE_NAME) == 0) {
/* platform disk utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, FS_ALARM_ID);
resource_config[i].mounted = MOUNTED;
resource_config[i].res_type = RESOURCE_TYPE__FILESYSTEM_USAGE ;
}
else if (strcmp(resource_config[i].resource, INSTANCE_RESOURCE_NAME) == 0) {
/* platform disk utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, INSTANCE_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__FILESYSTEM_USAGE ;
}
else if (strcmp(resource_config[i].resource, V_CINDER_THINPOOL_RESOURCE_NAME) == 0) {
/* platform virtual thin pool utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, V_CINDER_THINPOOL_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__FILESYSTEM_USAGE ;
}
else if (strcmp(resource_config[i].resource, V_NOVA_THINPOOL_RESOURCE_NAME) == 0) {
/* platform virtual thin pool utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, V_NOVA_THINPOOL_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__FILESYSTEM_USAGE ;
}
else if (strcmp(resource_config[i].resource, V_PORT_RESOURCE_NAME) == 0) {
/* vswitch port utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH,
V_PORT_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__PORT ;
}
else if (!strcmp(resource_config[i].resource, V_INTERFACE_RESOURCE_NAME) ||
!strcmp(resource_config[i].resource, V_LACP_INTERFACE_RESOURCE_NAME)) {
/* vswitch interface(lacp or otherwise) utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH,
V_INTERFACE_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__INTERFACE ;
}
else if (!strcmp(resource_config[i].resource, V_OVSDB_RESOURCE_NAME)) {
/* vswitch OVSDB manager utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH,
V_OVSDB_MANAGER_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__DATABASE_USAGE ;
}
else if (!strcmp(resource_config[i].resource, V_OPENFLOW_RESOURCE_NAME)) {
/* vswitch Openflow utilization */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH,
V_OPENFLOW_CONTROLLER_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__NETWORK_USAGE ;
}
else if (strcmp(resource_config[i].resource, REMOTE_LOGGING_RESOURCE_NAME) == 0) {
/* remote logging connectivity */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH,
REMOTE_LOGGING_CONTROLLER_CONNECTIVITY_ALARM_ID);
resource_config[i].res_type = RESOURCE_TYPE__CONNECTIVITY ;
}
else
{
resource_config[i].res_type = RESOURCE_TYPE__UNKNOWN ;
}
ilog ("Monitoring %2d: %s (%s)\n",
i,
resource_config[i].resource,
resource_config[i].severity);
mem_log_resource ( &resource_config[i] );
i++;
}
}
_rmon_ctrl_ptr->resources = i ;
ilog ("Monitoring %d Resources\n", _rmon_ctrl_ptr->resources );
return (PASS);
}
/*****************************************************************************
*
* Name : _inter_config_load
*
* Purpose : Load the content of each config file into interface_resource_config[x]
*
*****************************************************************************/
int _inter_config_load (void)
{
int i = 0 ;
for ( string_iter_ptr = interface_config_files.begin () ;
string_iter_ptr != interface_config_files.end () ;
string_iter_ptr++ )
{
if ( i >= MAX_RESOURCES )
{
wlog ("Cannot Monitor more than %d resources\n", MAX_RESOURCES );
break ;
}
/* Read the interface resource config file */
resource_config[i].mask = 0 ;
if (ini_parse( string_iter_ptr->data(), rmon_interface_config,
&interface_resource_config[i]) < 0)
{
ilog("Read Failure : %s\n", string_iter_ptr->data() );
}
else
{
dlog ("Config File : %s\n", string_iter_ptr->c_str());
ilog ("Monitoring %2d: %s (%s)\n", i, interface_resource_config[i].resource ,
interface_resource_config[i].severity );
interface_resource_config[i].i = i ;
interface_resource_config[i].failed = false ;
interface_resource_config[i].stage = RMON_STAGE__INIT ;
interface_resource_config[i].sev = SEVERITY_CLEARED ;
interface_resource_config[i].failed_send = 0;
interface_resource_config[i].alarm_raised = false;
/* add the alarm ids for the FM API per resource monitored */
if (strcmp(interface_resource_config[i].resource, OAM_INTERFACE_NAME) == 0) {
/* add the alarm id for the FM API per resource monitored */
snprintf(interface_resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, OAM_ALARM_ID);
snprintf(interface_resource_config[i].alarm_id_port, FM_MAX_BUFFER_LENGTH, OAM_PORT_ALARM_ID);
}
else if (strcmp(interface_resource_config[i].resource, MGMT_INTERFACE_NAME) == 0) {
/* add the alarm id for the FM API per resource monitored */
snprintf(interface_resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, MGMT_ALARM_ID);
snprintf(interface_resource_config[i].alarm_id_port, FM_MAX_BUFFER_LENGTH, MGMT_PORT_ALARM_ID);
}
else if (strcmp(interface_resource_config[i].resource, INFRA_INTERFACE_NAME) == 0) {
/* add the alarm id for the FM API per resource monitored */
snprintf(interface_resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, INFRA_ALARM_ID);
snprintf(interface_resource_config[i].alarm_id_port, FM_MAX_BUFFER_LENGTH, INFRA_PORT_ALARM_ID);
}
mem_log_interface_resource ( &interface_resource_config[i] );
i++;
}
}
_rmon_ctrl_ptr->interface_resources = i ;
ilog ("Monitoring %d Interface Resources\n", _rmon_ctrl_ptr->interface_resources );
return (PASS);
}
/*****************************************************************************
*
* Name : _thinmeta_config_load
*
* Purpose : Load the content of each config file into thinmeta_config[x]
*
*****************************************************************************/
int _thinmeta_config_load (void)
{
int i = 0 ;
/* Set hard-coded defaults for all structures */
for ( int j = 0; j < MAX_RESOURCES; j++)
{
thinmeta_resource_config_type * res;
res = &thinmeta_resource_config[i];
res->critical_threshold = THINMETA_DEFAULT_CRITICAL_THRESHOLD;
res->alarm_on = THINMETA_DEFAULT_ALARM_ON;
res->autoextend_on = THINMETA_DEFAULT_AUTOEXTEND_ON;
res->autoextend_by = THINMETA_DEFAULT_AUTOEXTEND_BY;
res->autoextend_percent = THINMETA_DEFAULT_AUTOEXTEND_PERCENT;
res->audit_period = THINMETA_DEFAULT_AUDIT_PERIOD;
}
/* Load resources */
for ( string_iter_ptr = config_files.begin () ;
string_iter_ptr != config_files.end () ;
string_iter_ptr++ )
{
if ( i >= MAX_RESOURCES )
{
wlog ("Cannot Monitor more than %d resources\n", MAX_RESOURCES );
break ;
}
/* Read the resource config file */
if (ini_parse( string_iter_ptr->data(), rmon_thinmeta_config,
&thinmeta_resource_config[i]) < 0)
{
ilog("Read Failure : %s\n", string_iter_ptr->data() );
}
else
{
thinmeta_resource_config_type * res;
res = &thinmeta_resource_config[i];
if (!res->section_exists)
{
dlog3 ("Config File : %s does not have a [%s] section\n",
string_iter_ptr->c_str(), THINMETA_CONFIG_SECTION);
continue;
}
dlog ("Config File : %s\n", string_iter_ptr->c_str());
/* validate loaded configuration */
if (!res->vg_name || !res->thinpool_name)
{
elog("Invalid VG and/or thinpool names for thinpool metadata "
"in config file: %s, disabling monitoring", string_iter_ptr->c_str());
res->critical_threshold = RESOURCE_DISABLE;
res->vg_name = THINMETA_INVALID_NAME;
res->thinpool_name = THINMETA_INVALID_NAME;
}
else if (res->critical_threshold > 99)
{
elog("Metadata monitoring error in config file: %s. Option critical_threshold > 99%%, "
"value in config file: %i, disabling monitoring",
string_iter_ptr->c_str(), res->critical_threshold)
res->critical_threshold = 0;
}
else if (res->alarm_on > 1)
{
elog("Metadata monitoring error in config file: %s. Option alarm_on is NOT boolean, "
"value in config file: %i, disabling monitoring", string_iter_ptr->c_str(), res->alarm_on);
res->critical_threshold = RESOURCE_DISABLE;
}
else if (res->autoextend_on > 1)
{
elog("Metadata monitoring error in config file: %s. Option autoextend_on is NOT boolean, "
"value in config file: %i, disabling monitoring",
string_iter_ptr->c_str(), res->autoextend_on)
res->critical_threshold = RESOURCE_DISABLE;
}
else if (res->autoextend_percent > 1)
{
elog("Metadata monitoring error in config file: %s. Option autoextend_percent is NOT boolean, "
"value in config file: %i, disabling monitoring",
string_iter_ptr->c_str(), res->autoextend_percent)
res->critical_threshold = RESOURCE_DISABLE;
}
else if ((res->autoextend_percent && res->autoextend_by > 100) ||
(res->autoextend_on && res->autoextend_by < 1))
{
elog("Metadata monitoring error in config file: %s. Option autoextend_by not in [1,100] interval, "
"value in config file: %i, disabling monitoring",
string_iter_ptr->c_str(), res->autoextend_by)
res->critical_threshold = RESOURCE_DISABLE;
}
else if ((res->audit_period < 1) || (res->audit_period > 10000))
{
elog("Metadata monitoring error in config file: %s. Option audit_period not in [1,10000] interval, "
"value in config file: %i, disabling monitoring",
string_iter_ptr->c_str(), res->audit_period)
res->critical_threshold = RESOURCE_DISABLE;
}
ilog ("%s/%s pool metadata monitored; resource index: %2d\n", res->vg_name ,
res->thinpool_name, i );
i++;
}
}
_rmon_ctrl_ptr->thinmeta_resources = i ;
ilog ("Monitoring %d Thinpool Metadata Resources\n", _rmon_ctrl_ptr->thinmeta_resources );
return (PASS);
}
/*****************************************************************************
*
* Name : rmon_hdlr_init
*
* Purpose : Init the handler but also support re-init that might occur over a SIGHUP
*
*****************************************************************************/
#define RMON_TIMER_TYPE__EVENT "event"
#define RMON_TIMER_TYPE__PM "pm"
#define RMON_TIMER_TYPE__NTP "ntp"
#define RMON_TIMER_TYPE__RES "resource"
#define RMON_TIMER_TYPE__THIN "thinpool"
int rmon_hdlr_init ( rmon_ctrl_type * ctrl_ptr )
{
/* Save the control pointer */
_rmon_ctrl_ptr = ctrl_ptr ;
mtcTimer_init ( rmonTimer_event, LOCALHOST, RMON_TIMER_TYPE__EVENT) ;
mtcTimer_init ( rmonTimer_pm, LOCALHOST, RMON_TIMER_TYPE__PM ) ;
if (is_controller())
mtcTimer_init ( rmonTimer_ntp,LOCALHOST, RMON_TIMER_TYPE__NTP ) ;
for ( int i = 0 ; i < MAX_RESOURCES ; i++ )
mtcTimer_init ( rtimer[i], LOCALHOST, RMON_TIMER_TYPE__RES );
ctrl_ptr->resources = 0 ;
for ( int i = 0 ; i < MAX_RESOURCES ; i++ )
mtcTimer_init ( thinmetatimer[i], LOCALHOST, RMON_TIMER_TYPE__THIN );
ctrl_ptr->thinmeta_resources = 0 ;
/* Initialize the Resource Monitor Array */
memset ( (char*)&resource_config[0], 0, sizeof(resource_config_type)*MAX_RESOURCES);
memset ( (char*)&interface_resource_config[0], 0, sizeof(interface_resource_config_type)*MAX_RESOURCES);
memset ( (char*)&thinmeta_resource_config[0], 0, sizeof(thinmeta_resource_config_type)*MAX_RESOURCES);
memset ( (char*)&registered_clt[0], 0, sizeof(registered_clients)*MAX_CLIENTS);
/* Read in the list of config files and their contents */
load_filenames_in_dir ( CONFIG_DIR, config_files ) ;
/* Read in the list of interface config files and their contents */
load_filenames_in_dir ( INT_CONFIG_DIR, interface_config_files ) ;
_thinmeta_config_load();
_config_files_load ();
// _inter_config_load ();
/* init Thin Metadata Monitoring after config reload - including timers */
thinmeta_init(thinmeta_resource_config, thinmetatimer, ctrl_ptr->thinmeta_resources);
/* Log the control setting going into the main loop */
mem_log_ctrl ( _rmon_ctrl_ptr );
/* Initialize instance mount monitoring */
if (pthread_mutex_init(&lock, NULL) != 0)
{
elog("mutex init failed \n");
}
t_data.thread_running = false;
t_data.resource_usage = MOUNTED;
t_data.nr_switches_count = 0;
t_data.pid = getpid();
return (PASS) ;
}
/*****************************************************************************
*
* Name : _set_severity
*
* Purpose : Restores the resource value and the severity of the alarm
*
*****************************************************************************/
void _set_resource_usage ( string reason_text, resource_config_type * ptr )
{
unsigned int found;
string res_val;
size_t last_index;
string temp_val;
char resource_usage[10];
/* extract the resource value from the reason text */
found = reason_text.find_last_of( ' ' );
temp_val = reason_text.substr(found+1);
last_index = temp_val.find_first_not_of("0123456789");
res_val = temp_val.substr(0, last_index);
snprintf (resource_usage, sizeof(resource_usage), res_val.c_str());
sscanf(resource_usage, "%lf", &ptr->resource_value);
}
/*****************************************************************************
*
* Name : build_entity_instance_id
*
* Purpose : build the alarm's entity_instance_id based on the
* resource type and alarm type.
*
*****************************************************************************/
void build_entity_instance_id ( resource_config_type *ptr, char *entity_instance_id )
{
dlog ("resource name: %s, resource type: %s, alarm type: %d \n", ptr->resource, ptr->type, ptr->alarm_type);
// Make certain the id is cleared
entity_instance_id[0] = 0;
if ( ptr->alarm_type == DYNAMIC_ALARM )
{
if ((ptr->type != NULL) && (strcmp(ptr->type, "lvg") == 0 ))
{
/* This case covers volume groups */
/* Use host=<x>.volumegroup=type for id*/
snprintf((char*)entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.volumegroup=%s", _rmon_ctrl_ptr->my_hostname, ptr->resource);
}
else
{
/* Use host=<x>.filesystem=type for id*/
snprintf(entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.filesystem=%s", _rmon_ctrl_ptr->my_hostname, ptr->resource);
}
}
else if ( ptr->alarm_type == STATIC_ALARM )
{
/* Use host=<x>.filesystem=type for id*/
snprintf(entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.filesystem=%s", _rmon_ctrl_ptr->my_hostname, ptr->resource);
}
else if ((ptr->alarm_type == STANDARD_ALARM) && (strstr(ptr->resource, V_MEMORY_RESOURCE_NAME) != NULL))
{
/* AVS memory */
snprintf(alarmData.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.processor=%d", _rmon_ctrl_ptr->my_hostname, ptr->socket_id);
}
else if (strstr(ptr->resource, V_CINDER_THINPOOL_RESOURCE_NAME) != NULL)
{
/* Cinder thin pool alarm should not be raised against a specific host */
/* as the volumes are synced between controllers through drbd. */
/* Instead we use a common entity instance id for both controllers. */
snprintf(entity_instance_id, FM_MAX_BUFFER_LENGTH, "host=controller");
}
else
{
/* Use hostname for alarm */
snprintf(entity_instance_id, FM_MAX_BUFFER_LENGTH, _rmon_ctrl_ptr->my_hostname);
}
dlog ("resource %s entity instance id: %s\n", ptr->resource, entity_instance_id);
return;
}
/*****************************************************************************
*
* Name : thinpool_virtual_space_usage_init
*
* Purpose : Determine if we should monitor virtual usage or not: no purpose
* in doing so if thin provisioning is not used.
*
* Params : index - the index of the virtual space resource
*
* Return : None.
*
*****************************************************************************/
void thinpool_virtual_space_usage_init(int index,
const char *poolName,
const char *poolOwner) {
if (!poolName or !poolOwner) {
slog ("No poolName or poolOwner provided");
return;
}
ilog("index = %d, poolName = %s, poolOwner = %s", index, poolName, poolOwner);
/* Buffer (and its size) for keeping the initial result after executing
the above command. */
char current_pool_type[BUFFER_SIZE];
const unsigned int buffer_size = BUFFER_SIZE;
/* The command for seeing if the pool type is thin. */
char lvm_thin_cmd[BUFFER_SIZE];
const char *thin_pool_expected_result = NULL;
MEMSET_ZERO(current_pool_type);
MEMSET_ZERO(lvm_thin_cmd);
if (strcmp(poolName, "nova-local-pool") == 0) {
const char *nova_thin_pool_expected_result = "thin-pool";
thin_pool_expected_result = nova_thin_pool_expected_result;
sprintf(lvm_thin_cmd, "lvs --segments | grep \"%s\" | awk '{print $5}'", poolName);
}
else if (strcmp(poolName, "cinder-volumes-pool") == 0) {
const char *cinder_thin_pool_expected_result = "thin";
thin_pool_expected_result = cinder_thin_pool_expected_result;
sprintf(lvm_thin_cmd, "cat /etc/cinder/cinder.conf | awk -F = '/^lvm_type.*=.*/ { print $2; }' | tail -n 1 | tr -d ' '");
}
else {
slog("Invalid pool name given.");
return;
}
/* Result code. */
int rc;
/* Execute the command. */
rc = execute_pipe_cmd(lvm_thin_cmd, current_pool_type, buffer_size);
/* If the command has been executed successfuly, continue. */
if (rc == PASS) {
if (current_pool_type != NULL) {
/* If the pool type is not thin, disable the alarm for virtual
usage. */
ilog("%s current pool type is set to = %s", poolOwner, current_pool_type);
if(strcmp(current_pool_type, thin_pool_expected_result) != 0) {
resource_config[index].alarm_status = ALARM_OFF;
ilog("%s LVM Thinpool Usage alarm off: thin provisioning not used", poolOwner);
} else {
resource_config[index].alarm_status = ALARM_ON;
ilog("%s LVM Thinpool Usage alarm on: thin provisioning used", poolOwner);
}
}
} else {
resource_config[index].alarm_status = ALARM_OFF;
elog("%s LVM Thinpool monitoring state unknown ; alarm disabled (rc:%i)",
poolOwner, rc);
}
}
/*****************************************************************************
*
* Name : virtual_space_usage_init
*
* Purpose : Determine if we should monitor virtual usage or not: no purpose
* in doing so if thin provisioning is not used.
*
* Return : None.
*
*****************************************************************************/
void virtual_space_usage_init(const char* resource_name) {
ilog ("Initialize thin pools for resource %s\n", resource_name);
int index;
if ( get_resource_index( resource_name, &index ) == PASS ) {
if (strcmp(resource_name, V_CINDER_THINPOOL_RESOURCE_NAME) == 0) {
thinpool_virtual_space_usage_init(index,"cinder-volumes-pool","Cinder");
} else if (strcmp(resource_name, V_NOVA_THINPOOL_RESOURCE_NAME) == 0) {
thinpool_virtual_space_usage_init(index, "nova-local-pool","Nova");
}
}
else {
wlog ("failed get_resource_index for resource %s\n", resource_name);
}
}
/*****************************************************************************
*
* Name : rmon_alarming_init
*
* Purpose : Clears any previously raised rmon alarms if rmon is restarted
*
*****************************************************************************/
void rmon_alarming_init ( resource_config_type * ptr )
{
dlog ("resource name: %s, resource type: %s, alarm type: %d \n", ptr->resource, ptr->type, ptr->alarm_type);
AlarmFilter alarmFilter;
SFmAlarmDataT *active_alarm = (SFmAlarmDataT*) calloc (1, sizeof (SFmAlarmDataT));
if (active_alarm == NULL)
{
elog("Failed to allocate memory for SFmAlarmDataT\n");
return;
}
build_entity_instance_id (ptr, alarmData.entity_instance_id);
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, ptr->alarm_id);
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, alarmData.entity_instance_id);
if (fm_get_fault( &alarmFilter, active_alarm) == FM_ERR_OK)
{
if (active_alarm != NULL) {
string reasonText(active_alarm->reason_text);
/* Set the resource severity */
ptr->failed = true;
ptr->alarm_raised = true;
ptr->count = ptr->num_tries;
if ( active_alarm->severity == FM_ALARM_SEVERITY_MINOR )
{
ptr->sev = SEVERITY_MINOR;
}
else if ( active_alarm->severity == FM_ALARM_SEVERITY_MAJOR )
{
ptr->sev = SEVERITY_MAJOR;
if ( ptr->res_type == RESOURCE_TYPE__FILESYSTEM_USAGE )
{
string err_res_name(ptr->resource);
_space_to_underscore(err_res_name);
/* clear host degrade for fs usage alarms */
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg), "%s %s:",
err_res_name.c_str(),
DEGRADE_CLEAR_MSG );
rmon_send_request ( ptr, _rmon_ctrl_ptr->clients );
}
}
else
{
ptr->sev = SEVERITY_CRITICAL;
}
resourceStageChange ( ptr, RMON_STAGE__MONITOR_WAIT );
if (strcmp(ptr->resource, INSTANCE_RESOURCE_NAME) != 0)
{
/* Set the resource severity */
_set_resource_usage( reasonText, ptr );
ilog ("%s setting previously failed resource alarm id: %s entity_instance_id: %s usage: %0.2f\n",
ptr->resource, ptr->alarm_id, alarmFilter.entity_instance_id, ptr->resource_value);
}
else
{
ilog ("%s setting previously failed resource alarm id: %s entity_instance_id: %s\n",
ptr->resource, ptr->alarm_id, alarmFilter.entity_instance_id);
}
}
}
free(active_alarm);
}
/*****************************************************************************
*
* Name : send_clear_msg
*
* Purpose : Send a message to all registered clients to set the node to
* available (clear the degrade)
*
*****************************************************************************/
void send_clear_msg ( int index )
{
int count = 0;
AlarmFilter alarmFilter;
SFmAlarmDataT *active_alarm = (SFmAlarmDataT*) calloc (1, sizeof (SFmAlarmDataT));
if (active_alarm == NULL)
{
elog("Failed to allocate memory for SFmAlarmDataT\n");
return;
}
string err_res_name(resource_config[index].resource);
_space_to_underscore(err_res_name);
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, resource_config[index].alarm_id);
build_entity_instance_id (&resource_config[index], alarmData.entity_instance_id);
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, alarmData.entity_instance_id);
/* Notify rmon clients of fault being cleared */
snprintf(resource_config[index].errorMsg, sizeof(resource_config[index].errorMsg),
"%s cleared_alarms_for_resource:", err_res_name.c_str());
/* check if there is an alarm first for this resource. If there is not then the node */
/* should not be in a degrade state */
EFmErrorT ret = fm_get_fault( &alarmFilter, active_alarm);
if ( (ret == FM_ERR_OK) && (active_alarm != NULL) )
{
while (( rmon_send_request ( &resource_config[index], _rmon_ctrl_ptr->clients ) != PASS ) && (count < 3 ))
{
wlog ("%s request send failed \n", resource_config[index].resource);
count++;
}
if (count > 2)
{
wlog ("%s request send failed, count:%d \n", resource_config[index].resource, count);
resource_config[index].failed_send++;
}
if ((resource_config[index].failed_send == MAX_FAIL_SEND) || (count < 3))
{
/* Reset the values to defaults */
swact_count = 0;
ilog("Setting resource: %s back to defaults \n", resource_config[index].resource);
resource_config[index].failed = false ;
resource_config[index].alarm_raised = false ;
resource_config[index].count = 0 ;
resource_config[index].sev = SEVERITY_CLEARED ;
resource_config[index].stage = RMON_STAGE__START ;
resource_config[index].failed_send = 0;
}
}
else //alarm not found or error
{
if (ret == FM_ERR_ENTITY_NOT_FOUND)
{
dlog ("Alarm not found for resource: %s entity_instance_id: %s \n", alarmFilter.alarm_id, alarmFilter.entity_instance_id);
}
else
{
wlog ("fm_get_fault failed for resource: %s entity_instance_id: %s err: %d\n", alarmFilter.alarm_id,
alarmFilter.entity_instance_id, ret);
}
if (active_alarm == NULL)
{
elog("fm_get_fault returned null active_alarm\n");
}
swact_count++;
if (swact_count == MAX_SWACT_COUNT)
{
/* Reset the values to defaults */
while (( rmon_send_request ( &resource_config[index], _rmon_ctrl_ptr->clients ) != PASS ) && (count < 3 ))
{
wlog ("%s request send failed \n", resource_config[index].resource);
count++;
}
swact_count = 0;
ilog("Setting resource: %s back to defaults \n", resource_config[index].resource);
resource_config[index].failed = false ;
resource_config[index].alarm_raised = false ;
resource_config[index].count = 0 ;
resource_config[index].sev = SEVERITY_CLEARED ;
resource_config[index].stage = RMON_STAGE__START ;
resource_config[index].failed_send = 0;
}
}
free(active_alarm);
}
/*****************************************************************************
*
* Name : read_fs_file
*
* Purpose : read the memory mapped dynamic file system file
*****************************************************************************/
void read_fs_file ( vector<string> & dynamic_resources )
{
FILE * pFile;
char buf[MAX_LEN];
int fd;
string delimiter = ",";
size_t pos;
string token;
struct stat fileInfo;
struct flock fl;
memset ((char *)&fileInfo, 0 , sizeof(fileInfo));
fl.l_whence = SEEK_SET;
fl.l_start = 0;
fl.l_len = 0;
fl.l_pid = getpid();
pFile = fopen (DYNAMIC_FS_FILE , "r");
if (pFile != NULL) {
fd = fileno(pFile);
/* lock the file */
fl.l_type = F_RDLCK;
/* lock the file for read and write */
fcntl(fd, F_SETLKW, &fl);
if (fd == -1)
{
elog("Error opening file for reading");
}
if (fstat(fd, &fileInfo) == -1)
{
elog("Error getting the file size");
}
char *map = static_cast<char*>( mmap(0, fileInfo.st_size, PROT_READ, MAP_SHARED, fd, 0));
if (map == MAP_FAILED)
{
elog("Error mmapping the file");
}
string str(map);
snprintf( buf, MAX_LEN, str.c_str());
/* free the mmapped memory */
if (munmap(map, fileInfo.st_size) == -1)
{
elog("Error un-mmapping the file");
}
fclose(pFile);
/* unlock the file */
fl.l_type = F_UNLCK;
fcntl(fd, F_SETLK, &fl);
while ((pos = str.find(delimiter)) != string::npos) {
/* separate the resources from the file */
token = str.substr(0, pos);
dynamic_resources.push_back(token);
dlog("reading resource %s \n", token.c_str());
str.erase(0, pos + delimiter.length());
}
}
}
/*****************************************************************************
*
* Name : add_dynamic_fs_resource
*
* Purpose : Add the dynamic file system resources
*****************************************************************************/
void add_dynamic_fs_resource ( bool send_response )
{
#ifdef WANT_FS_MONITORING
char resource[50];
char temp_resource[50];
char device [50];
char mount_point[50];
char temp_state[20];
char type [50];
char buf[200];
string criticality = "critical";
vector<string> resource_list;
int absolute_thresholds[3];
memset(absolute_thresholds, 0, sizeof(absolute_thresholds));
fs_index.clear();
fs_state.clear();
/* get a list of all the dynamic fs mounts */
read_fs_file(resource_list);
for(std::vector<string>::iterator it = resource_list.begin(); it != resource_list.end(); ++it)
{
string str = *it;
snprintf(buf, sizeof(buf), str.c_str());
// For resources without mounts the mount_point will be NULL
memset(&mount_point[0], 0, sizeof(mount_point));
sscanf(buf, "%49s %19s %49s %49s %49s", temp_resource, temp_state, type, device, mount_point);
string state(temp_state);
bool found = false;
if (mount_point[0] != '\0')
{
// for resources with mounts, the resource name is the mount value
snprintf(resource, FM_MAX_BUFFER_LENGTH, mount_point);
}
else
{
// for resources without mounts, the resource name is the device value
snprintf(resource, FM_MAX_BUFFER_LENGTH, device);
}
/* the dynamic file system is enabled, add it if need be */
for (int i=0; i<_rmon_ctrl_ptr->resources; i++)
{
if ( strcmp(resource, resource_config[i].resource) == 0)
{
dlog ("resource %s already exists, update the state to %s \n", resource, state.c_str());
/* resource already exists no need to add it again */
/* update the state, it may have changed */
fs_index.push_back(i);
fs_state.push_back(state);
found = true;
break;
}
}
if (!found) // new resource to monitor, lets add it
{
int enabled_resource = ALARM_OFF;
if (strcmp(temp_state,"enabled") == 0)
{
enabled_resource = ALARM_ON;
}
if (mount_point[0] != '\0')
{
save_fs_resource ( resource, criticality, enabled_resource, fs_percent, absolute_thresholds, DYNAMIC_ALARM, type, device, MOUNTED );
}
else
{
save_fs_resource ( resource, criticality, enabled_resource, fs_percent, absolute_thresholds, DYNAMIC_ALARM, type, device, NOT_MOUNTED );
}
if (enabled_resource == ALARM_ON) {
calculate_fs_usage( &resource_config[_rmon_ctrl_ptr->resources - 1] );
rmon_alarming_init( &resource_config[_rmon_ctrl_ptr->resources - 1] );
}
}
}
#endif
if (send_response)
{
#ifdef WANT_FS_MONITORING
ilog ("sending response to dynamic FS add, to the rmon client\n");
#else
ilog("dynamic filesystem monitoring moved to collectd\n");
#endif
/* let the rmon client know that we are done with the file */
rmon_resource_response(_rmon_ctrl_ptr->clients);
}
}
/*****************************************************************************
*
* Name : clear_alarm_for_resource
*
* Purpose : Clear the alarm of the resource passed in
*
*****************************************************************************/
void clear_alarm_for_resource ( resource_config_type * ptr )
{
dlog ("resource name: %s, resource type: %s, alarm type: %d \n", ptr->resource, ptr->type, ptr->alarm_type);
AlarmFilter alarmFilter;
build_entity_instance_id (ptr, alarmData.entity_instance_id);
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, ptr->alarm_id);
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, alarmData.entity_instance_id);
int ret = rmon_fm_clear(&alarmFilter);
if (ret == FM_ERR_OK)
{
ilog ("Cleared stale alarm %s for entity instance id: %s", alarmFilter.alarm_id, alarmFilter.entity_instance_id);
}
else if (ret == FM_ERR_ENTITY_NOT_FOUND)
{
dlog ("Stale alarm %s for entity instance id: %s was not found", alarmFilter.alarm_id, alarmFilter.entity_instance_id);
}
else
{
wlog ("Failed to clear stale alarm %s for entity instance id: %s error: %d", alarmFilter.alarm_id, alarmFilter.entity_instance_id, ret);
}
}
/*****************************************************************************
*
* Name : process_dynamic_fs_file
*
* Purpose : read the dynamic files directory and add the dynamic filesystem
* resources when the file is updated
*****************************************************************************/
void process_dynamic_fs_file()
{
int index = 0;
pthread_mutex_lock(&lock);
modifyingResources = true;
pthread_mutex_unlock(&lock);
add_dynamic_fs_resource(true);
pthread_mutex_lock(&lock);
modifyingResources = false;
pthread_mutex_unlock(&lock);
/* deal with changes of dynamic file system enabled state */
for (unsigned int i=0; i<fs_index.size(); i++)
{
index = fs_index.at(i);
if ( strcmp(fs_state.at(i).c_str(), "disable") == 0 )
{
/* resource has been disabled, stop alarming on it */
ilog("%s is no longer enabled\n", resource_config[index].resource);
if ( resource_config[index].failed == true )
{
resource_config[index].alarm_status = ALARM_OFF;
if ( _rmon_ctrl_ptr->clients > 0 )
{
//send a clear degrade node
send_clear_msg(index);
}
// we need to clear the resource's alarm if there was any set for this resource
clear_alarm_for_resource(&resource_config[index]);
}
else
{
/* There was no active alarm to clear */
ilog("Setting resource: %s back to defaults \n", resource_config[index].resource);
resource_config[index].alarm_status = ALARM_OFF;
resource_config[index].failed = false;
resource_config[index].alarm_raised = false;
resource_config[index].count = 0 ;
resource_config[index].sev = SEVERITY_CLEARED ;
resource_config[index].stage = RMON_STAGE__START ;
}
}
else if ( strcmp(fs_state.at(i).c_str(), "enabled") == 0 )
{
// resource has been enabled
if ( resource_config[index].alarm_status == ALARM_OFF )
{
/* Turn the resource checking back on if it was off */
resource_config[index].alarm_status = ALARM_ON;
//reset values
resource_config[index].failed = false;
resource_config[index].alarm_raised = false;
resource_config[index].count = 0 ;
resource_config[index].sev = SEVERITY_CLEARED ;
resource_config[index].stage = RMON_STAGE__START ;
rmon_alarming_init( &resource_config[index] );
ilog("%s is now enabled \n", resource_config[index].resource);
if (strcmp(resource_config[index].resource, CINDER_VOLUMES) == 0)
{
virtual_space_usage_init(V_CINDER_THINPOOL_RESOURCE_NAME);
}
if (strcmp(resource_config[index].resource, NOVA_LOCAL) == 0)
{
virtual_space_usage_init(V_NOVA_THINPOOL_RESOURCE_NAME);
}
}
else // alarm aready on (enabled)
{
ilog("%s is already enabled \n", resource_config[index].resource);
}
}
else
{
wlog("%s invalid dynamic file system state: %s \n", resource_config[index].resource, fs_state.at(i).c_str());
}
}
}
/*****************************************************************************
*
* Name : process_static_fs_file
*
* Purpose : Reads in the list of static file systems for monitoring
*
*****************************************************************************/
void process_static_fs_file()
{
FILE * pFile;
vector<string> mounts;
char buf[MAX_LEN];
char resource[50];
char type[50];
char device[50];
bool found = false;
int enabled_resource = ALARM_ON;
string criticality = "critical";
int absolute_thresholds[3] = {0};
pFile = fopen (STATIC_FS_FILE , "r");
if (pFile != NULL) {
ifstream fin( STATIC_FS_FILE );
string line;
while( getline( fin, line )) {
/* process each line */
mounts.push_back(line);
}
fclose(pFile);
for(std::vector<string>::iterator it = mounts.begin(); it != mounts.end(); ++it)
{
string str = *it;
snprintf(buf, MAX_LEN, str.c_str());
sscanf(buf, "%49s %49s %49s %d %d %d", resource, device, type, &absolute_thresholds[0], &absolute_thresholds[1], &absolute_thresholds[2]);
if (!found)
{
if (fs_percent == PERCENT_USED)
{
/* do not use the absolute thresholds */
memset(absolute_thresholds, 0, sizeof(absolute_thresholds));
}
/* add the resource */
save_fs_resource ( resource, criticality, enabled_resource, fs_percent, absolute_thresholds, STATIC_ALARM, type, device, MOUNTED );
calculate_fs_usage( &resource_config[_rmon_ctrl_ptr->resources - 1] );
}
}
}
else
{
elog("Error, no static file system file present at: %s\n", STATIC_FS_FILE);
}
}
/*****************************************************************************
*
* Name : rmon_timer_handler
*
* Purpose : Looks up the timer ID and asserts the corresponding ringer
*
*****************************************************************************/
void rmon_timer_handler ( int sig, siginfo_t *si, void *uc)
{
timer_t * tid_ptr = (void**)si->si_value.sival_ptr ;
/* Avoid compiler errors/warnings for parms we must
* have but currently do nothing with */
UNUSED(sig);
UNUSED(uc);
if ( !(*tid_ptr) )
{
// tlog ("Called with a NULL Timer ID\n");
return ;
}
/* is event rmon timer */
if ( *tid_ptr == rmonTimer_event.tid )
{
mtcTimer_stop_int_safe ( rmonTimer_event);
rmonTimer_event.ring = true ;
}
else if ( *tid_ptr == rmonTimer_pm.tid )
{
mtcTimer_stop_int_safe ( rmonTimer_pm);
rmonTimer_pm.ring = true ;
}
else if ( (is_controller()) && (*tid_ptr == rmonTimer_ntp.tid) )
{
mtcTimer_stop_int_safe ( rmonTimer_ntp);
rmonTimer_ntp.ring = true ;
}
else
{
bool found = false ;
for ( int i = 0 ; i < _rmon_ctrl_ptr->resources ; i++ )
{
if ( *tid_ptr == rtimer[i].tid )
{
mtcTimer_stop_int_safe ( rtimer[i] );
rtimer[i].ring = true ;
found = true ;
break ;
}
}
if ( !found )
{
for ( int i = 0 ; i < _rmon_ctrl_ptr->thinmeta_resources ; i++ )
{
if ( *tid_ptr == thinmetatimer[i].tid )
{
mtcTimer_stop_int_safe ( thinmetatimer[i] );
thinmetatimer[i].ring = true ;
found = true ;
break ;
}
}
}
if ( !found )
{
/* try and cleanup by stopping this unknown timer via its tid */
mtcTimer_stop_tid_int_safe (tid_ptr);
}
}
}
/*****************************************************************************
*
* Name : clear_ntp_alarms
*
* Purpose : Loop through each current alarms and deleted them if the server
* is now reachable or the server no longer is assigned to ntpq
*
*****************************************************************************/
void clear_ntp_alarms(std::list<string> &non_reachable_ntp_servers, unsigned int alarm_count, SFmAlarmDataT *active_alarms, bool clear_major_alarm)
{
dlog ("Total NTP alarm_count:%d", alarm_count);
AlarmFilter alarmFilter;
char alarm_to_search[FM_MAX_BUFFER_LENGTH];
fm_alarm_id alarm_id;
snprintf(alarm_id, FM_MAX_BUFFER_LENGTH, "%s", NTP_ALARM_ID);
// clear the major alarms if required
if (clear_major_alarm)
{
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, "%s", NTP_ALARM_ID );
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.ntp", _rmon_ctrl_ptr->my_hostname);
int ret = rmon_fm_clear(&alarmFilter);
if (ret != FM_ERR_OK)
{
if (ret != FM_ERR_ENTITY_NOT_FOUND)
{
wlog ("Failed to clear major alarm %s for entity instance id:%s error:%d", NTP_ALARM_ID, alarmFilter.entity_instance_id, ret);
}
}
else
{
ilog ("Cleared major alarm %s for entity instance id:%s", NTP_ALARM_ID, alarmFilter.entity_instance_id);
}
}
if (active_alarms == NULL)
{
elog ("Null pointer for active_alarms");
return;
}
// clear minor alarms if required
bool found;
std::list<string>::iterator iter;
std::list<string>::iterator iter_bad_list;
// for each NTP alarms in the system see if it match any of the invalid NTP servers
// if it does not match then the alarm must be removed since that NTP server
// is no longer being monitored or is now valid
for ( unsigned int i = 0; i < alarm_count; i++ )
{
if ( ((active_alarms+i)->severity) == FM_ALARM_SEVERITY_MINOR )
{
// Verify that this NTP minor alarm is still valid, This server could no longer exist or is now marked
// reachable
dlog ("Verify NTP minor alarm is still valid, entity instance id:%s", (active_alarms+i)->entity_instance_id);
found = false;
// check for stale minor alarm
for ( iter = non_reachable_ntp_servers.begin (); iter != non_reachable_ntp_servers.end (); iter++ )
{
// e.g. host=controller-0.ntp=102.111.2.2
snprintf(alarm_to_search, FM_MAX_BUFFER_LENGTH, "%s.ntp=%s", _rmon_ctrl_ptr->my_hostname, iter->c_str());
dlog ("Non reachable NTP server to search %s", iter->c_str());
if (strstr((active_alarms+i)->entity_instance_id, iter->c_str()) != NULL)
{
// server is in non reachable list, do not clear it
found = true;
dlog ("Alarm is still valid %s", iter->c_str());
break;
}
}
if (!found)
{
// lets clear it but only if it's this controller's alarm, it could be the peer controller's alarm
if (strstr((active_alarms+i)->entity_instance_id, _rmon_ctrl_ptr->my_hostname) != NULL)
{
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, "%s", NTP_ALARM_ID);
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s", (active_alarms+i)->entity_instance_id);
if (rmon_fm_clear(&alarmFilter) != FM_ERR_OK)
{
wlog ("Failed to clear minor alarm %s for entity instance id:%s", NTP_ALARM_ID, (active_alarms+i)->entity_instance_id);
}
else
{
ilog ("Cleared minor alarm %s for entity instance id:%s", NTP_ALARM_ID, (active_alarms+i)->entity_instance_id);
}
}
}
}
}
}
/*****************************************************************************
*
* Name : ntp_query_results
*
* Purpose : Analyze the return code from script query_ntp_servers.sh.
* Create alarms if the servers are non reachable, Clear alarms if they are
* now reachable
*
*****************************************************************************/
void ntp_query_results (int ntp_query_status )
{
dlog ("ntp_query_results ntp_query_status:%d", ntp_query_status);
std::list<string> non_reachable_ntp_servers;
// if no NTP servers are provisioned on the system, we still need to clear old NTP
// alarms if there are any. But we do not need to read the tmp server file.
if (ntp_query_status != NTP_NOT_PROVISIONED)
{
// read the temp file which contains a list of reachable and non reachable servers
// this file is the output from the query_ntp_servers.sh script
const char *server_info = "/tmp/ntpq_server_info";
FILE *pFile;
pFile = fopen(server_info, "r");
if (pFile != NULL)
{
const char * delim = ";\n\r";
char * ip;
char line[500];
int pos = 0;
while ( memset(line, 0, sizeof(line)) && (fgets((char*) &line, sizeof(line), pFile) != NULL) )
{
// the first line in the tmp file is the reachable servers, the second is the non reachable servers
if (pos == 1)
{
for (ip = strtok (line, delim); ip; ip = strtok (NULL, delim))
{
non_reachable_ntp_servers.push_back(ip);
dlog("Found non reachable NTP servers:%s\n", ip);
}
break;
}
pos++;
}
fclose(pFile);
}
else
{
elog("Failed to open file: %s\n", server_info);
return;
}
}
// retreive all the current NTP alarms
int rc;
unsigned int max_alarms=75;
fm_alarm_id alarm_id;
snprintf(alarm_id, FM_MAX_BUFFER_LENGTH, "%s", NTP_ALARM_ID);
SFmAlarmDataT *active_alarms = (SFmAlarmDataT*) calloc (max_alarms, sizeof (SFmAlarmDataT));
if (active_alarms == NULL)
{
elog ("Failed to allocate memory for NTP alarms");
return;
}
int ret = fm_get_faults_by_id( &alarm_id, active_alarms, &max_alarms);
if (!(ret == FM_ERR_OK || ret == FM_ERR_ENTITY_NOT_FOUND))
{
elog ("fm_get_faults_by_id failed trying to retreive all the NTP alarms, error:%d", ret);
free(active_alarms);
return;
}
// Clear alarms if required
bool clear_major_alarm = false;
bool created_major_alarm = false;
if ( ntp_query_status == NTP_NOT_PROVISIONED || ntp_query_status == NTP_SOME_REACHABLE || ntp_query_status == NTP_OK )
{
// We are going to clear the major alarm since there is at least one server selected or
// no servers are provisioned
clear_major_alarm = true;
}
// fm_get_faults_by_id returns the number of alarms found
if (max_alarms != 0)
{
// verify if alarms need to cleared and clear them
clear_ntp_alarms(non_reachable_ntp_servers, max_alarms, active_alarms, clear_major_alarm);
}
// There are no NTP servers provisioned so there is no alarms to raise
if (ntp_query_status == NTP_NOT_PROVISIONED)
{
return;
}
// Raise alarms if required
// Set up alarms data
AlarmFilter alarmFilter;
snprintf(alarmData.proposed_repair_action , sizeof(alarmData.proposed_repair_action), "Monitor and if condition persists, contact next level of support.");
snprintf(alarmData.alarm_id, FM_MAX_BUFFER_LENGTH, "%s", NTP_ALARM_ID);
strcpy(alarmData.uuid, "");
snprintf(alarmData.entity_type_id, FM_MAX_BUFFER_LENGTH, "ntp");
alarmData.alarm_state = FM_ALARM_STATE_SET;
alarmData.alarm_type = FM_ALARM_COMM;
alarmData.probable_cause = FM_ALARM_CAUSE_UNKNOWN;
alarmData.timestamp = 0;
alarmData.service_affecting = FM_FALSE;
alarmData.suppression = FM_FALSE;
// Here we raise the major alarm if required
if (ntp_query_status == NTP_NONE_REACHABLE || ntp_query_status == NTP_SOME_REACHABLE_NONE_SELECTED)
{
wlog("NTP configuration does not contain any valid or reachable NTP servers");
// Check if alarm is raised already
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.ntp", _rmon_ctrl_ptr->my_hostname);
bool found = false;
for ( unsigned int i = 0; i < max_alarms; i++ )
{
if ( strncmp((active_alarms+i)->entity_instance_id, alarmFilter.entity_instance_id, sizeof((active_alarms+i)->entity_instance_id)) == 0 )
{
// Alarm already exist
dlog("Alarm %s already raised for entity instance id:%s\n", NTP_ALARM_ID, alarmFilter.entity_instance_id);
found = true;
break;
}
}
// Alarm does not exist so raise it
if (!found && !created_major_alarm)
{
// Alarm does not exist so raise it
alarmData.severity = FM_ALARM_SEVERITY_MAJOR;
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text), "NTP configuration does not contain any valid or reachable NTP servers.");
snprintf(alarmData.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s", alarmFilter.entity_instance_id);
rc = rmon_fm_set(&alarmData, NULL);
if (rc == FM_ERR_OK )
{
ilog("Alarm %s created for entity instance id:%s \n", NTP_ALARM_ID, alarmData.entity_instance_id);
created_major_alarm = true;
}
else
{
ilog("Failed to create alarm %s for entity instance id:%s error: %d \n", NTP_ALARM_ID, alarmData.entity_instance_id, (int)rc);
}
}
}
// Here were raise alarms for individual servers
if (ntp_query_status != NTP_OK)
{
wlog("Some or all of the NTP servers are not reachable");
std::list<string>::iterator iter;
alarmData.severity = FM_ALARM_SEVERITY_MINOR;
// Loop through all the non reachable NTP servers
// Check to see if an alarms is lready raised for the server.
// If we do not find an alarm for the server then we raise it
for ( iter = non_reachable_ntp_servers.begin (); iter != non_reachable_ntp_servers.end (); iter++ )
{
bool found = false;
// Build the alarm entity instatance id
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s.ntp=%s", _rmon_ctrl_ptr->my_hostname, iter->c_str());
dlog("Search alarms for entity instance id:%s \n", alarmFilter.entity_instance_id);
for ( unsigned int i = 0; i < max_alarms; i++ )
{
if ( strncmp((active_alarms+i)->entity_instance_id, alarmFilter.entity_instance_id, sizeof((active_alarms+i)->entity_instance_id)) == 0 )
{
dlog("Alarm %s already raised for entity instance id:%s\n", NTP_ALARM_ID, alarmFilter.entity_instance_id);
found = true;
break;
}
}
// If the NTP alarm was not found then raise one for this NTP server
if (!found)
{
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text), "NTP address %s is not a valid or a reachable NTP server.", iter->c_str() );
snprintf(alarmData.entity_instance_id, FM_MAX_BUFFER_LENGTH, "%s", alarmFilter.entity_instance_id);
rc = rmon_fm_set(&alarmData, NULL);
if (rc == FM_ERR_OK )
{
ilog("Alarm %s created for entity instance id:%s \n", NTP_ALARM_ID, alarmData.entity_instance_id);
}
else
{
ilog("Failed to create alarm %s for entity instance id:%s error:%d \n", NTP_ALARM_ID, alarmData.entity_instance_id, (int)rc);
}
}
}
}
free(active_alarms);
return;
}
/*****************************************************************************
*
* Name : query_ntp_servers
*
* Purpose : execute script query_ntp_servers.sh which run the "ntpq -np"
* which query the healths of the NTP servers. The script will return a
* status code and also create a temporate file which will save the list
* of reachable and non reachable NTP servers. This temp file is required
* to generate proper alarms
*
*****************************************************************************/
int query_ntp_servers ( )
{
pid_t child_pid;
dlog ("Main Pid:%d \n", getpid() );
ntp_child_pid = child_pid = fork ();
if (child_pid == 0)
{
dlog ("Child Pid:%d \n", getpid() );
char* argv[] = {(char*)NTPQ_QUERY_SCRIPT, NULL};
char cmd[MAX_FILE_SIZE] ;
memset (cmd,0,MAX_FILE_SIZE);
snprintf ( &cmd[0], MAX_FILE_SIZE, "%s/%s", RMON_FILES_DIR, NTPQ_QUERY_SCRIPT );
bool close_file_descriptors = true ;
if ( setup_child ( close_file_descriptors ) != PASS )
{
exit(NTP_ERROR);
}
/* Set child to ignore child exit */
signal (SIGCHLD, SIG_DFL);
/* Setup the exec arguement */
int res = execv(cmd, argv);
elog ( "Failed to run %s return code:%d error:%s\n", cmd, res, strerror(errno) );
exit (NTP_ERROR);
}
if ( child_pid == -1 )
{
elog ("Fork failed (%s)\n", strerror(errno));
/* TODO: Consider making this a critical fault
* after 100 retries.
* All possibilities based on man page are
* due to resource limitations and if that does
* not resolve in 100 retries then ip probably will never.
**/
return (FAIL);
}
return (PASS);
}
/*****************************************************************************
*
* Name : rmonHdlr_ceilometer_handler
*
* Purpose : Handles the ceilometer sample create response message
*
*****************************************************************************/
void rmonHdlr_ceilometer_handler( struct evhttp_request *req, void *arg )
{
if ( !req )
{
elog (" Request Timeout\n");
ceilometerEvent.status = FAIL_TIMEOUT;
goto _ceilometer_handler_done ;
}
ceilometerEvent.status = rmonHttpUtil_status(ceilometerEvent);
if ( ceilometerEvent.status != PASS )
{
elog ("ceilometer HTTP request Failed (%d)\n", ceilometerEvent.status);
rmonHttpUtil_get_response(ceilometerEvent);
goto _ceilometer_handler_done ;
}
_ceilometer_handler_done:
event_base_loopbreak((struct event_base *)arg);
}
/*****************************************************************************
*
* Name : generate_ceilometer_pm
*
* Purpose : Generate ceilometer PMs through the REST API
*
*****************************************************************************/
void generate_ceilometer_pm ( string r_id, string m_id, string m_type,
string m_unit, string m_volume,
string m_metadata )
{
int rc = PASS;
daemon_config_type * cfg_ptr = daemon_get_cfg_ptr();
string command_path="";
string host_ip = cfg_ptr->keystone_auth_host;
int port = cfg_ptr->ceilometer_port;
int count = 0;
rmonHttpUtil_libEvent_init ( &ceilometerEvent, CEILOMETER_EVENT_SIG, host_ip, port);
ceilometerEvent.address.append("/v2/meters/");
ceilometerEvent.address.append(m_id);
ceilometerEvent.user_agent = "ceilometerclient.openstack.common.apiclient";
ceilometerEvent.payload = "[{";
ceilometerEvent.payload.append("\"resource_id\":\"");
ceilometerEvent.payload.append(r_id);
ceilometerEvent.payload.append("\",\"counter_name\":\"");
ceilometerEvent.payload.append(m_id);
ceilometerEvent.payload.append("\",\"counter_type\":\"");
ceilometerEvent.payload.append(m_type);
ceilometerEvent.payload.append("\",\"counter_unit\":\"");
ceilometerEvent.payload.append(m_unit);
ceilometerEvent.payload.append("\",\"counter_volume\":\"");
ceilometerEvent.payload.append(m_volume);
ceilometerEvent.payload.append("\",\"resource_metadata\":");
// the resource metadata is dictionary of key-value pairs
ceilometerEvent.payload.append(m_metadata);
ceilometerEvent.payload.append("}]");
dlog ("Payload is : %s\n", ceilometerEvent.payload.c_str());
rc = rmonHttpUtil_api_request (CEILOMETER_SAMPLE_CREATE, ceilometerEvent, command_path);
do
{
if ( rc != PASS )
{
count++;
wlog ("ceilometer failed request (%d) ... retrying (%d)\n", rc, count);
}
rmonHttpUtil_log_event (ceilometerEvent);
} while ( ( rc!=PASS ) && ( count < REST_API_RETRY_COUNT ) );
if ( rc!= PASS )
{
elog ("ceilometer sample create Failed (%d) (cnt:%d)\n", rc, count);
}
}
void clear_rmon_api_counts ( registered_clients * ptr )
{
if ( ptr->b2b_miss_count > ptr->b2b_miss_peak )
{
ptr->b2b_miss_peak = ptr->b2b_miss_count ;
}
if ( ptr->mesg_err_cnt > ptr->mesg_err_peak )
{
ptr->mesg_err_peak = ptr->mesg_err_cnt ;
}
ptr->b2b_miss_count = 0 ;
ptr->send_err_cnt = 0 ;
ptr->recv_err_cnt = 0 ;
ptr->mesg_err_cnt = 0 ;
}
/*****************************************************************************
*
* Name : _space_to_underscore
*
* Purpose : Converts spaces in a string to underscores
* *****************************************************************************/
void _space_to_underscore (string & str )
{
char space = ' ';
for(unsigned int i = 0; i < str.size(); i++)
{
if(str[i] == space)
{
str[i] = '_';
}
}
}
/*****************************************************************************
*
* Name : set_alarm_defaults
*
* Purpose : Set the defaults for the fm alarms
* *****************************************************************************/
void set_alarm_defaults ( resource_config_type * ptr )
{
strcpy(alarmData.uuid, "");
/* common data for all alarm messages */
snprintf(alarmData.entity_type_id, FM_MAX_BUFFER_LENGTH, "system.host");
build_entity_instance_id (ptr, alarmData.entity_instance_id);
alarmData.alarm_state = FM_ALARM_STATE_SET;
alarmData.alarm_type = FM_ALARM_OPERATIONAL;
alarmData.probable_cause = FM_ALARM_THRESHOLD_CROSSED;
alarmData.timestamp = 0;
alarmData.service_affecting = FM_FALSE;
alarmData.suppression = FM_TRUE;
snprintf(alarmData.alarm_id, FM_MAX_BUFFER_LENGTH, ptr->alarm_id);
}
/*****************************************************************************
*
* Name : resource_handler
*
* Purpose : Handle the failed resources and raise alarms through
* the FM API as well as calling a function to notify registered clients
*****************************************************************************/
int resource_handler ( resource_config_type * ptr )
{
int rc = RETRY ;
AlarmFilter alarmFilter;
string err_res_name(ptr->resource);
_space_to_underscore(err_res_name);
if ( ptr->stage < RMON_STAGE__STAGES )
{
dlog2 ("%s %s Stage %d\n", ptr->resource, rmonStages_str[ptr->stage], ptr->stage );
}
else
{
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
switch ( ptr->stage )
{
case RMON_STAGE__START:
{
dlog ( "%s failed:%d set_cnt:%d debounce_cnt:%d\n",
ptr->resource,
ptr->failed,
ptr->count,
ptr->debounce_cnt);
break ;
}
case RMON_STAGE__MANAGE:
{
/* send messages to maintnance in thresholds are crossed */
if (ptr->alarm_status == ALARM_ON)
{
/* set up the fm api alarm defaults */
set_alarm_defaults( ptr );
if ( strcmp(ptr->resource, MEMORY_RESOURCE_NAME) == 0 )
{
snprintf(alarmData.proposed_repair_action , sizeof(alarmData.proposed_repair_action),
"Monitor and if condition persists, contact next level of support; may require additional memory on Host.");
}
else if ( strcmp(ptr->resource, INSTANCE_RESOURCE_NAME) == 0 )
{
snprintf(alarmData.proposed_repair_action , sizeof(alarmData.proposed_repair_action),
"Check Management and Infrastructure Networks and Controller or Storage Nodes.");
}
else
{
if ((ptr->type != NULL) && (strcmp(ptr->type, "lvg") == 0 ))
{
snprintf(alarmData.proposed_repair_action , sizeof(alarmData.proposed_repair_action),
"Monitor and if condition persists, consider adding additional physical volumes to the volume group.");
}
else
{
snprintf(alarmData.proposed_repair_action , sizeof(alarmData.proposed_repair_action),
"Monitor and if condition persists, contact next level of support.");
}
}
if ( ptr->sev == SEVERITY_MINOR )
{
alarmData.severity = FM_ALARM_SEVERITY_MINOR;
if ( ptr->percent == PERCENT_USED ) {
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->resource, ptr->minor_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->resource, ptr->minor_threshold, ptr->resource_value);
}
else {
ilog ("Filesystem threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->minor_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->minor_threshold, ptr->resource_value);
}
} else {
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->resource, ptr->minor_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->resource, ptr->minor_threshold_abs_node0, ptr->resource_value);
} else {
ilog ("Filesystem threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->minor_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->minor_threshold_abs_node0, ptr->resource_value);
}
}
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg),
"%s minor_threshold_set", err_res_name.c_str());
}
else if ( ptr->sev == SEVERITY_MAJOR )
{
alarmData.severity = FM_ALARM_SEVERITY_MAJOR;
if (strcmp(ptr->resource, INSTANCE_RESOURCE_NAME) != 0)
{
if (ptr->percent == PERCENT_USED){
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->resource, ptr->major_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->resource, ptr->major_threshold, ptr->resource_value);
}
else {
ilog ("Filesystem threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->major_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem threshold exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->major_threshold, ptr->resource_value);
}
} else {
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->resource, ptr->major_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->resource, ptr->major_threshold_abs_node0, ptr->resource_value);
} else {
ilog ("Filesystem threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->major_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->major_threshold_abs_node0, ptr->resource_value);
}
}
}
else if (strcmp(ptr->resource, INSTANCE_RESOURCE_NAME) == 0)
{
/* instance alarming is a special case of alarm */
wlog ("No access to remote VM volumes.\n");
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"No access to remote VM volumes.");
}
if ( ptr->res_type == RESOURCE_TYPE__FILESYSTEM_USAGE )
{
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg),
"%s %s",err_res_name.c_str(), DEGRADE_CLEAR_MSG );
}
else
{
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg),
"%s major_threshold_set",err_res_name.c_str());
}
}
else if ( ptr->sev == SEVERITY_CRITICAL )
{
alarmData.severity = FM_ALARM_SEVERITY_CRITICAL;
if (ptr->percent == PERCENT_USED){
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->resource, ptr->critical_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->resource, ptr->critical_threshold, ptr->resource_value);
}
else {
ilog ("Filesystem threshold exceeded; threshold: %d%%, actual: %.2f%%. \n",
ptr->critical_threshold, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem threshold exceeded; threshold: %u%%, actual: %.2f%%.",
ptr->critical_threshold, ptr->resource_value);
}
} else {
if ( ptr->alarm_type == STANDARD_ALARM )
{
ilog ("%s threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->resource, ptr->critical_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"%s threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->resource, ptr->critical_threshold_abs_node0, ptr->resource_value);
} else {
ilog ("Filesystem threshold exceeded; threshold: %dMB, remaining value: %.2fMB. \n",
ptr->critical_threshold_abs_node0, ptr->resource_value);
snprintf(alarmData.reason_text, sizeof(alarmData.reason_text),
"Filesystem threshold exceeded; threshold: %uMB, remaining value: %.2fMB.",
ptr->critical_threshold_abs_node0, ptr->resource_value);
}
}
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg),
"%s major_threshold_set",err_res_name.c_str());
}
rc = rmon_fm_set(&alarmData, NULL);
if (rc == FM_ERR_OK ) {
ilog("%s: %s alarm\n",
ptr->resource,
FmAlarmSeverity_to_string(alarmData.severity).c_str());
ptr->alarm_raised = true;
} else {
ilog("%s: %s alarm failed (rc:%d)\n",
ptr->resource,
FmAlarmSeverity_to_string(alarmData.severity).c_str(),
(int)rc);
}
if (ptr->alarm_raised)
{
if ((_rmon_ctrl_ptr->clients > 0) && (ptr->failed_send < MAX_FAIL_SEND))
{
/* If degrade debounce is non-zero then this
* alarm condition is candidate for host degrade */
if (ptr->debounce)
{
if ( rmon_send_request ( ptr, _rmon_ctrl_ptr->clients ) != PASS )
{
ptr->failed_send++;
wlog ("%s request send failed (count:%d)\n",
ptr->resource,
ptr->failed_send );
}
else
{
ptr->failed_send = 0;
}
}
}
else
{
ptr->failed_send = 0;
}
resourceStageChange ( ptr, RMON_STAGE__MONITOR_WAIT );
}
}
else {
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
break;
}
case RMON_STAGE__IGNORE:
{
//nothing to do here, go to the finished stage
resourceStageChange ( ptr, RMON_STAGE__FINISH );
break ;
}
case RMON_STAGE__MONITOR_WAIT:
{
if ((_rmon_ctrl_ptr->clients > 0) && (ptr->failed_send < MAX_FAIL_SEND) && (ptr->failed_send > 0))
{
if ( rmon_send_request ( ptr, _rmon_ctrl_ptr->clients ) != PASS )
{
wlog ("%s request send failed \n", ptr->resource);
ptr->failed_send++;
}
else
{
ptr->failed_send = 0;
}
}
break;
}
case RMON_STAGE__FINISH:
{
if ((ptr->alarm_status == ALARM_ON) && (ptr->alarm_raised))
{
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, ptr->alarm_id);
build_entity_instance_id (ptr, alarmData.entity_instance_id);
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, alarmData.entity_instance_id);
ilog ("%s alarm clear\n", ptr->resource );
/* clear the alarm */
EFmErrorT ret = rmon_fm_clear(&alarmFilter);
if (( ret == FM_ERR_OK ) || ( ret == FM_ERR_ENTITY_NOT_FOUND ))
{
if (ret == FM_ERR_ENTITY_NOT_FOUND)
{
dlog ("%s alarm clear failed, entity '%s' not found",
ptr->resource, alarmData.entity_instance_id);
}
snprintf(ptr->errorMsg, sizeof(ptr->errorMsg), "%s cleared_alarms_for_resource", err_res_name.c_str());
if ( (_rmon_ctrl_ptr->clients > 0) && ( ptr->failed_send < MAX_FAIL_SEND ) && (ret == FM_ERR_OK) )
{
while (( rmon_send_request ( ptr, _rmon_ctrl_ptr->clients ) != PASS ) &&
( ptr->failed_send < MAX_FAIL_SEND ))
{
wlog ("%s request send failed \n", ptr->resource);
ptr->failed_send++;
}
ptr->alarm_raised = false;
ptr->failed_send = 0;
ptr->failed = false ;
ptr->count = 0 ;
ptr->sev = SEVERITY_CLEARED ;
ptr->stage = RMON_STAGE__START ;
}
else
{
ptr->alarm_raised = false;
ptr->failed_send = 0;
ptr->failed = false ;
ptr->count = 0 ;
ptr->sev = SEVERITY_CLEARED ;
ptr->stage = RMON_STAGE__START ;
}
}
else
{
wlog("%s alarm clear failed, entity '%s' (rc:%d)\n",
ptr->resource,
alarmData.entity_instance_id,
ret);
}
}
else
{
ptr->alarm_raised = false;
ptr->failed_send = 0;
ptr->failed = false ;
ptr->count = 0 ;
ptr->sev = SEVERITY_CLEARED ;
ptr->stage = RMON_STAGE__START ;
}
rc = PASS ;
break ;
}
default:
{
slog ("%s Invalid stage (%d)\n", ptr->resource, ptr->stage );
/* Default to finish for invalid case.
* If there is an issue then it will be detected */
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
}
return rc;
}
/*****************************************************************************
*
* Name : process_failures
*
* Purpose : Check whether a percentage resource is to be failed or a failure
* threshold is to be cleared by the resource_handler
*
*****************************************************************************/
void process_failures ( resource_config_type * ptr )
{
if (ptr->stage == RMON_STAGE__INIT)
{
/* first time after restart/reboot, clear the alarm if the first reading is good */
resourceStageChange ( ptr, RMON_STAGE__START );
if (ptr->resource_value < ptr->minor_threshold)
{
// assuming we left as alarm on last time
ptr->alarm_status = ALARM_ON;
ptr->alarm_raised = true;
ptr->failed = true;
ilog("%s Setting the state to FINISH\n", ptr->resource);
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
// Now we start counting as normal ...
}
else
{
if (ptr->failed)
{
/* If the resource is already failed, check to see if it is to be cleared */
if ((( ptr->sev == SEVERITY_MINOR) && ( ptr->resource_value < ptr->minor_threshold )) ||
(( ptr->sev == SEVERITY_MAJOR) && ( ptr->resource_value < ptr->major_threshold )) ||
(( ptr->sev == SEVERITY_CRITICAL) && ( ptr->resource_value < ptr->critical_threshold )))
{
if (ptr->count > ptr->num_tries)
ptr->count = ptr->num_tries;
if (ptr->count > 0)
ptr->count--;
if (ptr->count == 0) {
ptr->sev = SEVERITY_CLEARED;
ilog("%s Setting the state to FINISH\n", ptr->resource);
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
}
else
{
/* While in failed state, the resource usage must sustain normal level
* num_tries number of times before an alarm can be cleared. Keep incrementing the counter
* as it will be set to num_tries in the above block as soon as resource usage returns to
* normal level.*/
ptr->count++;
// rmon needs to send degrade assert message periodically as the
// condition might be cleared by maintenance over controller swact.
//
// added meaning to the debounce config setting.
// must be non-zero to degrade the host.
if ((ptr->alarm_raised) && (ptr->debounce) &&
(_rmon_ctrl_ptr->clients > 0))
{
if ( rmon_send_request ( ptr, _rmon_ctrl_ptr->clients ) != PASS )
{
ptr->failed_send++ ;
wlog ("%s request send failed (count:%d)\n",
ptr->resource,
ptr->failed_send);
}
else
{
mlog ("%s rmon_send_request ok\n", ptr->resource );
ptr->failed_send = 0 ;
}
}
else
{
/* typical path for resources that
* - do not degrade host
* - do not raise alarms */
dlog ("%s: alarm:%d debounce:%d clients:%d\n",
ptr->resource,
(ptr->alarm_raised),
(ptr->debounce),
(_rmon_ctrl_ptr->clients));
}
}
}
}
/* Check to see if a resource is over the failure thresholds for: minor, major and critical failures */
if (( ptr->resource_value >= ptr->minor_threshold ) &&
( ptr->resource_value < ptr->major_threshold )
&& (ptr->sev != SEVERITY_MINOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries) {
ptr->failed = true;
ptr->sev = SEVERITY_MINOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value >= ptr->major_threshold ) &&
( ptr->resource_value < ptr->critical_threshold )
&& (ptr->sev != SEVERITY_MAJOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_MAJOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value >= ptr->critical_threshold )&&
(ptr->sev != SEVERITY_CRITICAL))
{
ptr->count++;
if (ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_CRITICAL;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else
{
/* if the host experienced a resource blip in the previous audit run and usage
* is now back at the normal level, decrement the count.*/
if ((!ptr->failed) && (ptr->count > 0)){
ptr->count--;
dlog("Resource %s is back at the normal level, count is set to %d", ptr->resource, ptr->count);
}
}
}
/*****************************************************************************
*
* Name : process_failures_absolute
*
* Purpose : Check whether an absolute resource is to be failed or a
* failure threshold is to be cleared by the resource_handler
*
*****************************************************************************/
void process_failures_absolute ( resource_config_type * ptr )
{
int node = 0;
if (strcmp(ptr->resource,"processor_node1") == 0)
{
/* per node memory checking is enabled */
node = 1;
}
if (ptr->failed) {
/* If the resource is already failed, check to see if it is to be cleared */
if (node == 0) {
if ((( ptr->sev == SEVERITY_MINOR) && ( ptr->resource_value > ptr->minor_threshold_abs_node0 )) ||
(( ptr->sev == SEVERITY_MAJOR) && ( ptr->resource_value > ptr->major_threshold_abs_node0 )) ||
(( ptr->sev == SEVERITY_CRITICAL) && ( ptr->resource_value > ptr->critical_threshold_abs_node0 )))
{
if (ptr->count > ptr->num_tries)
ptr->count = ptr->num_tries;
if (ptr->count > 0)
ptr->count--;
if (ptr->count == 0) {
ptr->sev = SEVERITY_CLEARED;
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
}
else
{
/* While in failed state, the resource usage must sustain normal level
* num_tries number of times before an alarm can be cleared. Keep incrementing the counter
* as it will be set to num_tries in the above block as soon as resource usage returns to
* normal level.*/
ptr->count++;
}
}
else {
if ((( ptr->sev == SEVERITY_MINOR) && ( ptr->resource_value > ptr->minor_threshold_abs_node1 )) ||
(( ptr->sev == SEVERITY_MAJOR) && ( ptr->resource_value > ptr->major_threshold_abs_node1 )) ||
(( ptr->sev == SEVERITY_CRITICAL) && ( ptr->resource_value > ptr->critical_threshold_abs_node1 )))
{
if (ptr->count > ptr->num_tries)
ptr->count = ptr->num_tries;
if (ptr->count > 0)
ptr->count--;
if (ptr->count == 0) {
ptr->sev = SEVERITY_CLEARED;
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
}
else
{
/* While in failed state, the resource usage must sustain normal level
* num_tries number of times before an alarm can be cleared. Keep incrementing the counter
* as it will be set to num_tries in the above block as soon as resource usage returns to
* normal level.*/
ptr->count++;
}
}
}
if (node == 0) {
/* Check to see if a resource is over the failure thresholds for: minor, major and critical failures node 0 */
if (( ptr->resource_value <= ptr->minor_threshold_abs_node0 ) &&
( ptr->resource_value > ptr->major_threshold_abs_node0 ) &&
(ptr->sev != SEVERITY_MINOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_MINOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value <= ptr->major_threshold_abs_node0 ) &&
( ptr->resource_value > ptr->critical_threshold_abs_node0 ) &&
(ptr->sev != SEVERITY_MAJOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_MAJOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value < ptr->critical_threshold_abs_node0 )&&
(ptr->sev != SEVERITY_CRITICAL))
{
ptr->count++;
if (ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_CRITICAL;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else
{
/* if the host experienced a resource blip in the previous audit run and usage
* is now back at the normal level, decrement the count.*/
if ((!ptr->failed) && (ptr->count > 0)){
ptr->count--;
dlog("Resource %s is back at the normal level, count is set to %d", ptr->resource, ptr->count);
}
}
} else {
/* Check to see if a resource is over the failure thresholds for: minor, major and critical failures node 1 */
if (( ptr->resource_value <= ptr->minor_threshold_abs_node1 ) &&
( ptr->resource_value > ptr->major_threshold_abs_node1 ) &&
(ptr->sev != SEVERITY_MINOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_MINOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value <= ptr->major_threshold_abs_node1 ) &&
( ptr->resource_value > ptr->critical_threshold_abs_node1 ) &&
(ptr->sev != SEVERITY_MAJOR))
{
ptr->count++;
if ( ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_MAJOR;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else if (( ptr->resource_value < ptr->critical_threshold_abs_node1 )&&
(ptr->sev != SEVERITY_CRITICAL))
{
ptr->count++;
if (ptr->count >= ptr->num_tries){
ptr->failed = true;
ptr->sev = SEVERITY_CRITICAL;
resourceStageChange ( ptr, RMON_STAGE__MANAGE);
}
}
else
{
/* if the host experienced a resource blip in the previous audit run and usage
* is now back at the normal level, decrement the count.*/
if ((!ptr->failed) && (ptr->count > 0)){
ptr->count--;
dlog("Resource %s is back at the normal level, count is set to %d", ptr->resource, ptr->count);
}
}
}
}
void update_total_clients (int total_clients)
{
_rmon_ctrl_ptr->clients = total_clients;
}
void add_registered_client (registered_clients client)
{
registered_clt[_rmon_ctrl_ptr->clients] = client;
ilog("added registered client: %s \n", client.client_name);
}
/*****************************************************************************
*
* Name : add_fs_resource
*
* Purpose : Add a dynamic or static fs resource by reading
* the: /etc/rmonfiles.d/dynamic.conf file
*****************************************************************************/
void add_fs_resource ( int resource_index, int criticality_index, int enabled,
int percent, int abs_values[3], int alarm_type,
int types_index, int devices_index, int mounted )
{
int fs_resource_index;
get_resource_index( FS_RESOURCE_NAME, &fs_resource_index );
int i = _rmon_ctrl_ptr->resources;
if (i >= MAX_RESOURCES) {
wlog ("Cannot Monitor more than %d resources\n", MAX_RESOURCES );
}
else {
resource_config[i].resource = dynamic_resource.at(resource_index).c_str();
resource_config[i].severity = criticality_resource.at(criticality_index).c_str();
resource_config[i].type = types.at(types_index).c_str();
resource_config[i].device = devices.at(devices_index).c_str();
resource_config[i].critical_threshold = UNUSED_CRITICAL; // initialization
resource_config[i].critical_threshold_abs_node0 = UNUSED_CRITICAL_ABS_NODE0;
resource_config[i].num_tries = DEFAULT_NUM_TRIES;
resource_config[i].alarm_status = enabled;
resource_config[i].percent = percent;
resource_config[i].mounted = mounted;
resource_config[i].alarm_type = alarm_type;
resource_config[i].debounce = resource_config[fs_resource_index].debounce;
// percentage based threshold measure
switch (percent) {
case PERCENT_USED:
if (abs_values[0] == 0) {
// if this is a static mounted file system resource
// then use common threshold values provided for the
// File System Resource
if ( (alarm_type == STATIC_ALARM) && (mounted == MOUNTED) ) {
resource_config[i].minor_threshold =
resource_config[fs_resource_index].minor_threshold;
resource_config[i].major_threshold =
resource_config[fs_resource_index].major_threshold;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold =
resource_config[fs_resource_index].critical_threshold;
}
resource_config[i].num_tries =
resource_config[fs_resource_index].num_tries;
}
else {
/* There are no specific percent thresholds for
the dynamic resource, use defaults */
resource_config[i].minor_threshold = FS_MINOR;
resource_config[i].major_threshold = FS_MAJOR;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold = FS_CRITICAL;
}
}
}
else if (abs_values[0] != 0) {
/* Specific percent thresholds are defined for the dynamic resource */
resource_config[i].minor_threshold = abs_values[0];
resource_config[i].major_threshold = abs_values[1];
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold = abs_values[2];
}
}
break;
case PERCENT_UNUSED:
if (abs_values[0] == 0) {
// if this is a static mounted file system then use common
// threshold values provided for the File System Resource
if ( (alarm_type == STATIC_ALARM) && (mounted == MOUNTED) ) {
resource_config[i].minor_threshold_abs_node0 =
resource_config[fs_resource_index].minor_threshold_abs_node0;
resource_config[i].major_threshold_abs_node0 =
resource_config[fs_resource_index].major_threshold_abs_node0;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold_abs_node0 = DEFAULT_CRITICAL_ABS_NODE0;
}
resource_config[i].num_tries =
resource_config[fs_resource_index].num_tries;
}
else {
/* If the percent thresholds are selected
* use the default thresholds for the absolute
* value thresholds for the dynamic resource */
resource_config[i].minor_threshold_abs_node0 = DEFAULT_MINOR_ABS_NODE0;
resource_config[i].major_threshold_abs_node0 = DEFAULT_MAJOR_ABS_NODE0;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold_abs_node0 = DEFAULT_CRITICAL_ABS_NODE0;
}
}
}
else if (abs_values[0] != 0) {
/* Specific absolute value thresholds are specified for the dynamic resource */
resource_config[i].minor_threshold_abs_node0 = abs_values[0];
resource_config[i].major_threshold_abs_node0 = abs_values[1];
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold_abs_node0 = abs_values[2];
}
}
break;
}
ilog ("Monitoring %2d: %-20s (%s) (%s)\n", i, resource_config[i].resource ,
resource_config[i].severity, (enabled ? "enabled" : "disabled") );
/* Init the timer for this resource */
mtcTimer_init ( rtimer[i] ) ;
rtimer[i].hostname = "localhost" ;
rtimer[i].service = resource_config[i].resource ;
resource_config[i].i = i;
resource_config[i].failed = false ;
resource_config[i].count = 0 ;
resource_config[i].stage = RMON_STAGE__START ;
resource_config[i].sev = SEVERITY_CLEARED ;
resource_config[i].failed_send = 0;
resource_config[i].alarm_raised = false;
resource_config[i].res_type = RESOURCE_TYPE__FILESYSTEM_USAGE ;
/* add the alarm id for the FM API per resource monitored */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, FS_ALARM_ID);
mem_log_resource ( &resource_config[i] );
i++;
_rmon_ctrl_ptr->resources = i;
}
}
/*****************************************************************************
*
* Name : save_dynamic_resource
*
* Purpose : Loops through resources and only adds a dynamic file system
* resource if it does not yet exist
******************************************************************************/
void save_fs_resource ( string resource_name, string criticality,
int enabled, int percent,
int abs_values[3], int alarm_type,
string type, string device, int mounted)
{
size_t resource_index;
size_t criticality_index;
size_t types_index;
size_t devices_index;
bool newResource = true;
for (int k=0; k< _rmon_ctrl_ptr->resources; k++) {
if (strcmp(resource_config[k].resource, resource_name.c_str()) == 0) {
newResource = false;
break;
}
}
if (newResource == true) {
dlog ("%s(%s) fs resource add in %s state\n", resource_name.c_str(),
criticality.c_str(), (enabled) ? "enabled" : "disabled");
dynamic_resource.push_back(resource_name);
resource_index = dynamic_resource.size() - 1;
/* add the criticality value to a vector for permenant storage */
criticality_resource.push_back(criticality);
criticality_index = criticality_resource.size() - 1;
types.push_back(type);
types_index = types.size() - 1;
devices.push_back(device);
devices_index = devices.size() - 1;
add_fs_resource ( resource_index, criticality_index, enabled, percent, abs_values, alarm_type, types_index, devices_index, mounted );
}
}
/*****************************************************************************
*
* Name : add_dynamic_mem_resource
*
* Purpose : Add a dynamic memory resource at runtime based on the name and criticality.
* The resource has both custom or default percent and absolute thresholds.
* *****************************************************************************/
int add_dynamic_mem_resource ( int resource_index, int criticality_index,
double r_value, int percent, int abs_values[3],
const char * alarm_id, int socket_id=0 )
{
int i = _rmon_ctrl_ptr->resources;
int new_index = i;
if (i >= MAX_RESOURCES) {
wlog ("Cannot Monitor more than %d resources\n", MAX_RESOURCES );
}
else {
resource_config[i].resource = dynamic_resource.at(resource_index).c_str();
resource_config[i].severity = criticality_resource.at(criticality_index).c_str();
if ((percent == 1) && (abs_values[0] == 0)) {
/* There are no specific percent thresholds for the dynamic resource, use defaults */
resource_config[i].minor_threshold = DEFAULT_MINOR;
resource_config[i].major_threshold = DEFAULT_MAJOR;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold = DEFAULT_CRITICAL;
} else {
resource_config[i].critical_threshold = UNUSED_CRITICAL;
}
}
else if ((percent == 1) && (abs_values[0] != 0)) {
/* Specific percent thresholds are defined for the dynamic resource */
resource_config[i].minor_threshold = abs_values[0];
resource_config[i].major_threshold = abs_values[1];
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold = abs_values[2];
} else {
resource_config[i].critical_threshold = UNUSED_CRITICAL;
}
}
if ((percent == 0) && (abs_values[0] == 0)) {
/* If the percent thresholds are selected use the default thresholds for the absolute
* value thresholds for the dynamic resource */
resource_config[i].minor_threshold_abs_node0 = DEFAULT_MINOR_ABS_NODE0;
resource_config[i].major_threshold_abs_node0 = DEFAULT_MAJOR_ABS_NODE0;
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold_abs_node0 = DEFAULT_CRITICAL_ABS_NODE0;
} else {
resource_config[i].critical_threshold_abs_node0 = UNUSED_CRITICAL_ABS_NODE0;
}
resource_config[i].minor_threshold_abs_node1 = DEFAULT_MINOR_ABS_NODE1;
resource_config[i].major_threshold_abs_node1 = DEFAULT_MAJOR_ABS_NODE1;
resource_config[i].critical_threshold_abs_node1 = DEFAULT_CRITICAL_ABS_NODE1;
}
else if ((percent == 0) && (abs_values[0] != 0)) {
/* Specific absolute value thresholds are specified for the dynamic resource */
resource_config[i].minor_threshold_abs_node0 = abs_values[0];
resource_config[i].major_threshold_abs_node0 = abs_values[1];
if (_rmon_ctrl_ptr->rmon_critical_thr == 1) {
resource_config[i].critical_threshold_abs_node0 = abs_values[2];
} else {
resource_config[i].critical_threshold_abs_node0 = UNUSED_CRITICAL_ABS_NODE0;
}
resource_config[i].minor_threshold_abs_node1 = DEFAULT_MINOR_ABS_NODE1;
resource_config[i].major_threshold_abs_node1 = DEFAULT_MAJOR_ABS_NODE1;
resource_config[i].critical_threshold_abs_node1 = DEFAULT_CRITICAL_ABS_NODE1;
}
resource_config[i].num_tries = DEFAULT_NUM_TRIES;
resource_config[i].alarm_status = DEFAULT_ALARM_STATUS;
resource_config[i].percent = percent;
ilog ("Monitoring %2d: Dynamic Resource- %s (%s)\n", i, resource_config[i].resource ,
resource_config[i].severity );
/* Init the timer for this resource */
mtcTimer_init ( rtimer[i] ) ;
rtimer[i].hostname = "localhost" ;
rtimer[i].service = resource_config[i].resource ;
resource_config[i].i = i;
resource_config[i].failed = false ;
resource_config[i].count = 0 ;
resource_config[i].resource_value = r_value ;
resource_config[i].resource_prev = r_value ;
resource_config[i].stage = RMON_STAGE__START ;
resource_config[i].sev = SEVERITY_CLEARED ;
resource_config[i].alarm_type = STANDARD_ALARM;
resource_config[i].failed_send = 0;
resource_config[i].alarm_raised = false;
resource_config[i].socket_id = socket_id;
/* add the alarm id for the FM API per resource monitored */
snprintf(resource_config[i].alarm_id, FM_MAX_BUFFER_LENGTH, alarm_id);
mem_log_resource ( &resource_config[i] );
i++;
_rmon_ctrl_ptr->resources = i;
}
return new_index;
}
/*****************************************************************************
*
* Name : save_dynamic_mem_resource
*
* Purpose : Loops through resources and only adds a memory resource if it does not yet
* exist
******************************************************************************/
int save_dynamic_mem_resource ( string resource_name, string criticality,
double r_value, int percent, int abs_values[3],
const char * alarm_id, int socket_id=0 )
{
size_t resource_index;
size_t criticality_index;
bool newResource = true;
int updated_index;
for (int k=0; k< _rmon_ctrl_ptr->resources; k++) {
if (strcmp(resource_config[k].resource, resource_name.c_str()) == 0) {
resource_config[k].resource_value=
resource_config[k].resource_prev = r_value;
updated_index = k;
newResource = false;
break;
}
}
if (newResource == true) {
dynamic_resource.push_back(resource_name);
resource_index = dynamic_resource.size() - 1;
/* add the criticality value to a vector for permenant storage */
criticality_resource.push_back(criticality);
criticality_index = criticality_resource.size() - 1;
updated_index = add_dynamic_mem_resource(resource_index, criticality_index,
r_value, percent, abs_values,
alarm_id, socket_id);
rmon_alarming_init( &resource_config[updated_index] );
resource_config[updated_index].resource_prev =
resource_config[updated_index].resource_value= r_value;
}
return updated_index;
}
/*****************************************************************************
*
* Name : calculate_fs_usage
*
* Purpose : Calculate the file system usage as a percentage or an absolute value
* for the number of MiB remaining overall and in a specific fs. The calculation
* is done by executing the df command and getting the response for each type
* of filesystem being monitored.
*****************************************************************************/
void calculate_fs_usage ( resource_config_type * ptr )
{
dlog("%s, is mounted resource: %d is enabled: %d\n", ptr->resource, ptr->mounted, ptr->alarm_status);
FILE *pFile;
int last_index;
char fsLine[128];
char buf[200];
double fsUsage = 0;
char mounted_on[50], file_system[50], capacity[10];
unsigned long long size, used, available;
string res_val;
double cap_percent;
double MiB = 1024.0;
double free_units = 0;
double usage_percents = 0;
double total_units = 0;
if (ptr->mounted == MOUNTED)
{
if (strcmp(ptr->resource, FS_RESOURCE_NAME) == 0)
{
// We do not calculate the total for filesystem
// Resource FS_RESOURCE_NAME represents the total filesystem
return;
}
else
{
snprintf(buf, sizeof(buf), "timeout 2 df -T -P --local %s 2>/dev/null", ptr->resource);
}
/* convert output of "df -P" from KiB to MiB */
if(!(pFile = popen(buf, "r")))
{
elog("Error, command df is not executed on resource: %s\n", ptr->resource);
}
else
{
while (memset(fsLine, 0, sizeof(fsLine)) && (fgets((char*) &fsLine, sizeof(fsLine), pFile) != NULL))
{
sscanf(fsLine, "%49s %*s %llu %llu %llu %9s %49s", file_system, &size, &used, &available, capacity, mounted_on);
if (strcmp(mounted_on, ptr->resource) == 0)
{
string temp_val(capacity);
// exclude percentage (%) sign
last_index = temp_val.find_first_not_of("0123456789");
res_val = temp_val.substr(0, last_index);
snprintf(capacity, sizeof(capacity), res_val.c_str());
sscanf(capacity, "%lf", &cap_percent);
if (ptr->percent == PERCENT_USED)
{
fsUsage = cap_percent;
ptr->resource_value = fsUsage;
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ) )
{
plog("filesystem: %s usage: %.2f%%\n",
ptr->resource, ptr->resource_value);
}
}
else
{
fsUsage = (double) (((100 - cap_percent) / 100) * size);
fsUsage = fsUsage / MiB;
ptr->resource_value = fsUsage;
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ) )
{
plog("filesystem: %s has %f (MiB) (free)\n",
ptr->resource, ptr->resource_value);
}
}
// The size of the file system is 2X the user specified size to allow upgrades.
// Currently we are alarming on the used size but instead the alarming should be based on used size /2.
// As a result there is no indication to the user that they have may have eaten into the reserved space
// for upgrades resulting in an aborted upgrade.
if (strcmp(mounted_on, POSTGRESQL_FS_PATH) == 0)
{
ptr->resource_value = ptr->resource_value / 2;
}
}
}
}
pclose(pFile);
}
else if(strcmp(ptr->resource, NOVA_LOCAL) == 0)
{
/*rmon queries the thin pool usage if the volume group is nova-local*/
snprintf(buf, sizeof(buf), "timeout 2 lvdisplay -C --noheadings --nosuffix -o data_percent --units m "
"/dev/nova-local/nova-local-pool 2>/dev/null");
if(!(pFile = popen(buf, "r")))
{
elog("Error, command lvdisplay free units is not executed \n");
}
else
{
while (memset(fsLine, 0, sizeof(fsLine)) && (fgets((char*) &fsLine, sizeof(fsLine), pFile) != NULL))
{
usage_percents = atof(fsLine);
}
pclose(pFile);
}
ptr->resource_value = usage_percents;
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("filesystem: %s, usage: %f%% \n", ptr->resource, ptr->resource_value);
}
}
else if(strcmp(ptr->resource, CINDER_VOLUMES) == 0)
{
/*rmon queries the thin pool usage if the volume group is cinder-volumes*/
snprintf(buf, sizeof(buf), "timeout 2 lvdisplay -C --noheadings --nosuffix -o data_percent --units m "
"/dev/cinder-volumes/cinder-volumes-pool 2>/dev/null");
if(!(pFile = popen(buf, "r")))
{
elog("Error, command lvdisplay free units is not executed \n");
}
else
{
while (memset(fsLine, 0, sizeof(fsLine)) && (fgets((char*) &fsLine, sizeof(fsLine), pFile) != NULL))
{
usage_percents = atof(fsLine);
}
pclose(pFile);
}
ptr->resource_value = usage_percents;
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("filesystem: %s, usage: %.2f%% \n", ptr->resource, ptr->resource_value);
}
}
else
{
/* for the unmounted dynamic file system resources, use the vgdisplay command to get vg free units */
snprintf(buf, sizeof(buf), "timeout 2 vgdisplay -C --noheadings --nosuffix -o vg_free --units m %s 2>/dev/null", ptr->resource);
if(!(pFile = popen(buf, "r")))
{
elog("Error, command vgdisplay free units is not executed \n");
}
else
{
while (memset(fsLine, 0, sizeof(fsLine)) && (fgets((char*) &fsLine, sizeof(fsLine), pFile) != NULL))
{
free_units = atof(fsLine);
}
pclose(pFile);
}
/* for the unmounted dynamic file system resources, use the vgdisplay command to get vg size */
snprintf(buf, sizeof(buf), "timeout 2 vgdisplay -C --noheadings --nosuffix -o vg_size --units m %s 2>/dev/null", ptr->resource );
if(!(pFile = popen(buf, "r")))
{
elog("Error, command vgdisplay total units is not executed \n");
}
else
{
while (memset(fsLine, 0, sizeof(fsLine)) && (fgets((char*) &fsLine, sizeof(fsLine), pFile) != NULL))
{
total_units = atof(fsLine);
}
pclose(pFile);
}
if ( ptr->percent == PERCENT_USED )
{
if (total_units != 0)
{
ptr->resource_value = (double) (( (total_units - free_units) / total_units ) * 100);
}
else
{
ptr->resource_value = 0;
}
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("volume-group: %s, usage: %.2f%%\n", ptr->resource, ptr->resource_value);
}
}
else
{
ptr->resource_value = free_units;
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("volume-group: %s, %.2f (MiB) free\n", ptr->resource, ptr->resource_value);
}
}
}
}
/*****************************************************************************
*
* Name : init_memory_checking
*
* Purpose : Get the memory accounting used either 0: overcommit or 1: strict
*****************************************************************************/
void init_memory_accounting()
{
const char *strict_memory_file = "/proc/sys/vm/overcommit_memory";
ifstream mem_file ( strict_memory_file );
string strict_line;
if (mem_file.is_open())
{
while ( getline (mem_file, strict_line) ) {
IS_STRICT = atoi(strict_line.c_str());
}
mem_file.close();
}
}
/*****************************************************************************
*
* Name : thinpool_calcVirtUsage
*
* Purpose : Obtain the percentage of the used virtual space in thin
* provisioning.
*
* Params : index - the index of the monitored resource (virtual space)
*
* Return : PASS/FAIL
*
*****************************************************************************/
int thinpool_calcVirtUsage(int index,
const char *poolName,
const char *poolOwner,
const char *allocParam) {
/* Initialize the variables used in calculating the virtual usage. */
double provisioned_capacity = 0;
double total_capacity = 0;
double allocation_ratio = 1;
double ratio = 0;
double MiB = 1024.0;
/* Buffer (and its size) for keeping the initial result after executing
the above commands. */
char result[BUFFER_SIZE];
const unsigned int buffer_size = BUFFER_SIZE;
/* Return code. */
int rc;
/* Save the necessary commands for obtaining the information about virtual
thin pool usage: provisioned capacity, total capacity and maximum
oversubscription ratio. */
const char *provisioned_capacity_cmd = NULL;
const char *allocation_ratio_cmd = NULL;
char total_capacity_cmd[BUFFER_SIZE];
snprintf(total_capacity_cmd, sizeof(total_capacity_cmd),
"lvs --units m --segments | grep \"%s\" | awk '{print $6}' | sed '$s/.$//'",
poolName);
if (strcmp (poolOwner, "Cinder") == 0) {
const char *cinder_provisioned_capacity_cmd ="lvs --units m | grep \"volume-[.]*\" | awk '{ sum+=$4} END {print sum}'";
const char *cinder_allocation_ratio_cmd = "cat /etc/cinder/cinder.conf | grep \"^max_over_subscription_ratio\" | cut -d '=' -f 2";
provisioned_capacity_cmd = cinder_provisioned_capacity_cmd;
allocation_ratio_cmd = cinder_allocation_ratio_cmd;
} else if (strcmp (poolOwner, "Nova") == 0) {
const char *nova_provisioned_capacity_cmd = "lvs --units m | grep \"[.]*_disk\" | awk '{ sum+=$4} END {print sum}'";
provisioned_capacity_cmd = nova_provisioned_capacity_cmd;
}
/* Determine the provisioned capacity. */
rc = execute_pipe_cmd(provisioned_capacity_cmd, result, buffer_size);
if (rc != PASS) {
wlog("%s LVM Thinpool ; unable to query provisioned capacity (rc:%i)",
poolOwner, rc);
return (FAIL);
}
provisioned_capacity = atof(result);
dlog("%s LVM Thinpool provisioned capacity is %f", poolOwner, provisioned_capacity);
/* If the threshold is of percentage type, then also determine the total
thin pool capacity and the max oversubscription ratio. */
rc = execute_pipe_cmd(total_capacity_cmd, result, buffer_size);
if (rc != PASS) {
elog("%s LVM Thinpool ; unable to query total capacity (rc:%i)",
poolOwner, rc);
return (FAIL);
}
total_capacity = atof(result);
dlog("%s LVM Thinpool total capacity is %f",
poolOwner, total_capacity);
if (strcmp (poolOwner, "Cinder") == 0) {
rc = execute_pipe_cmd(allocation_ratio_cmd, result, buffer_size);
if (rc != PASS) {
elog("%s LVM Thinpool %s ratio could not be determined (rc:%i)",
allocParam, poolOwner, rc);
return (FAIL);
}
allocation_ratio = atof(result);
} else if (strcmp (poolOwner, "Nova") == 0) {
allocation_ratio = 1.0;
}
dlog("%s LVM Thinpool %s is %f", poolOwner, allocParam, allocation_ratio);
/* If the allocation_ratio is 0 or hasn't been found, its default
value should be 1. */
if (allocation_ratio == 0)
allocation_ratio = 1;
/* Compute the current virtual space usage of the thin pool. */
if (total_capacity != 0){
ratio = provisioned_capacity / (total_capacity * allocation_ratio) * 100;
} else {
/*3 minutes (30 sec * rate_throttle = 180 sec)*/
/* Change the warning log to a debug log to avoid generating this log in
rmond.log when Cinder is Ceph backended. Once the repackaging of cinder_virtual_resource.conf
and nova_virtual_resource.conf is done, we will change it back to warning log. */
dlog("%s LVM Thinpool total capacity is 0\n", poolOwner);
return (FAIL);
}
/* Update the resource value configuration. */
if (resource_config[index].percent == 1) {
resource_config[index].resource_value = ratio;
if ( log_value ( resource_config[index].resource_value,
resource_config[index].resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("%s LVM Thinpool Usage: %.2f%%", poolOwner, ratio);
}
}
else {
resource_config[index].resource_value =
((total_capacity * allocation_ratio) - provisioned_capacity) * MiB;
if ( log_value ( resource_config[index].resource_value,
resource_config[index].resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("%s LVM Thinpool has %.2f (MiB) free",
poolOwner,
resource_config[index].resource_value);
}
}
return (PASS);
}
/*****************************************************************************
*
* Name : calculate_virtual_space_usage
*
* Purpose : Obtain the percentage of the used virtual space in thin
* provisioning.
*
* Params : index - the index of the monitored resource (virtual space)
*
* Return : PASS/FAIL
*
*****************************************************************************/
int calculate_virtual_space_usage(int index, const char* constant) {
int rc = 0;
if (strcmp(constant, V_CINDER_THINPOOL_RESOURCE_NAME) == 0) {
rc = thinpool_calcVirtUsage(index,
"cinder-volumes-pool",
"Cinder",
"max_over_subscription_ratio");
} else if (strcmp(constant, V_NOVA_THINPOOL_RESOURCE_NAME) == 0) {
rc = thinpool_calcVirtUsage(index,
"nova-local-pool",
"Nova",
"disk_allocation_ratio");
}
return rc;
}
/*****************************************************************************
*
* Name : calculate_memory_usage
*
* Purpose : Calculate the memory usage as a percentage or absolute value for the
* number of MiB left. The overall average memory usage as well as the per NUMA
* node memory usage is computed.
*****************************************************************************/
void calculate_memory_usage( int index ) {
const char *mem_info = "/proc/meminfo";
FILE *pFile;
char memoryLine[40];
char attribute_name[30];
double memUsage, memUsageHuge;
char *line0 = &memoryLine[0];
char *line3 = &memoryLine[3];
char *line10 = &memoryLine[10];
unsigned long int value;
unsigned long int avail = 0;
unsigned long int memTotal;
int resource_name_size = 100;
string resource_name_huge = "processor_hugepages_";
string resource_name = "processor_";
char numa_node[resource_name_size];
string criticality = "critical";
double MiB = 1024.0;
int absolute_thresholds[3];
memoryinfo memInfo;
struct dirent *ent;
DIR *numa_node_dir;
vector<string> numa_files;
vector<string> node_files;
memset ( (char*)&memInfo, 0, sizeof(memoryinfo));
if ((pFile = fopen(mem_info, "r")) == NULL){
dlog("failed to open: /proc/meminfo \n");
}
else {
while (memset(memoryLine, 0, sizeof(memoryLine)) && (fgets((char*) &memoryLine, sizeof(memoryLine), pFile) != NULL)) {
if (*line3 == 'T') {
/* match MemTotal */
value = 0UL;
if (sscanf(memoryLine, "MemTotal: %lu", &value) == 1) {
memInfo.MemTotal = value;
continue;
}
} else if (*line3 == 'F') {
/* match MemFree */
value = 0UL;
if (sscanf(memoryLine, "MemFree: %lu", &value) == 1) {
memInfo.MemFree = value;
continue;
}
} else if (*line3 == 'f') {
/* match Buffers */
value = 0UL;
if (sscanf(memoryLine, "Buffers: %lu", &value) == 1) {
memInfo.Buffers = value;
continue;
}
} else if (*line3 == 'h') {
/* match Cached */
value = 0UL;
if (sscanf(memoryLine, "Cached: %lu", &value) == 1) {
memInfo.Cached = value;
continue;
}
} else if ((*line0 == 'S') && (*line3 == 'c')) {
/* match Slab Reclaimable */
value = 0UL;
if (sscanf(memoryLine, "SReclaimable: %lu", &value) == 1) {
memInfo.SlabReclaimable = value;
continue;
}
} else if ((*line0 == 'C') && (*line10 == 't')) {
/* match CommitLimit */
value = 0UL;
if (sscanf(memoryLine, "CommitLimit: %lu", &value) == 1) {
memInfo.CommitLimit = value;
continue;
}
} else if ((*line0 == 'C') && (*line10 == 'A')) {
/* match Committed_AS */
value = 0UL;
if (sscanf(memoryLine, "Committed_AS: %lu", &value) == 1) {
memInfo.Committed_AS = value;
continue;
}
} else if ((*line0 == 'H') && (*line10 == 'T')) {
/* match Hugepages_Total */
value = 0UL;
if (sscanf(memoryLine, "HugePages_Total: %lu", &value) == 1) {
memInfo.HugePages_Total = value;
continue;
}
}
else if ((*line0 == 'H') && (*line10 == 'z')) {
/* match Hugepagesize */
value = 0UL;
if (sscanf(memoryLine, "Hugepagesize: %lu", &value) == 1) {
memInfo.Hugepagesize = value;
continue;
}
}
else if ((*line0 == 'A') && (*line3 == 'n')) {
/* match AnonPages */
value = 0UL;
if (sscanf(memoryLine, "AnonPages: %lu", &value) == 1) {
memInfo.AnonPages = value;
continue;
}
}
}
fclose(pFile);
}
avail = memInfo.MemFree + memInfo.Buffers + memInfo.Cached + memInfo.SlabReclaimable;
memTotal = avail + memInfo.AnonPages;
dlog("memTotal: %lu\n", memTotal);
/* average memory utilization */
if (IS_STRICT == 1) {
/* strict memory checking enabled */
if (resource_config[index].percent == 1) {
memUsage = (double) memInfo.Committed_AS / memInfo.CommitLimit;
memUsage = memUsage * 100;
} else {
memUsage = (double) (memInfo.CommitLimit - memInfo.Committed_AS) / MiB;
}
} else {
if (resource_config[index].percent == 1)
{
memUsage = (double) memInfo.AnonPages / memTotal;
memUsage = memUsage * 100;
} else
{
memUsage = (double) avail / MiB;
}
}
resource_config[index].resource_value = memUsage;
if (resource_config[index].percent == 1)
{
if ( log_value ( resource_config[index].resource_value,
resource_config[index].resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("%s: %.2f%%\n",
resource_config[index].resource, memUsage);
}
}
else
{
if ( log_value ( resource_config[index].resource_value,
resource_config[index].resource_prev,
DEFAULT_LOG_VALUE_STEP ))
{
plog("%s: %.2f (MiB) free\n",
resource_config[index].resource, memUsage);
}
}
if ((numa_node_dir= opendir ("/sys/devices/system/node/")) != NULL) {
/* print all the files and directories within directory */
while ((ent = readdir (numa_node_dir)) != NULL) {
if (strstr(ent->d_name, "node") != NULL) {
numa_files.push_back(ent->d_name);
}
}
closedir (numa_node_dir);
}
/* loop through all NUMA nodes to get memory usage per NUMA node */
for (unsigned int p=0; p<numa_files.size(); p++) {
snprintf(numa_node, resource_name_size, "/sys/devices/system/node/%s/meminfo",numa_files.at(p).c_str());
resource_name += numa_files.at(p);
resource_name_huge += numa_files.at(p);
pFile = fopen (numa_node, "r");
while (memset(memoryLine, 0, sizeof(memoryLine)) && (fgets((char*) &memoryLine, sizeof(memoryLine), pFile) != NULL)) {
sscanf(memoryLine, "Node %*d %29s %lu", attribute_name, &value);
if (strstr(attribute_name, "MemTotal") != NULL) {
/* match MemTotal */
memInfo.MemTotal = value;
} else if (strstr(attribute_name, "MemFree") != NULL) {
/* match MemFree */
memInfo.MemFree = value;
} else if (strstr(attribute_name, "FilePages") != NULL) {
/* match FilePages */
memInfo.FilePages = value;
} else if (strstr(attribute_name, "SReclaimable") != NULL) {
/* match SReclaimable */
memInfo.SlabReclaimable = value;
} else if (strstr(attribute_name, "HugePages_Total") != NULL) {
/* match HugePages_Total */
memInfo.HugePages_Total = value;
} else if (strstr(attribute_name, "HugePages_Free") != NULL) {
/* match HugePages_Free */
memInfo.HugePages_Free = value;
} else if (strstr(attribute_name, "AnonPages") != NULL) {
/* match AnonPages ) */
memInfo.AnonPages = value;
}
}
fclose(pFile);
if (_rmon_ctrl_ptr->per_node == 1) {
/* if set to 1 get the per NUMA node memory values */
memset(absolute_thresholds, 0, sizeof(absolute_thresholds));
avail = memInfo.MemFree + memInfo.FilePages + memInfo.SlabReclaimable;
memTotal = avail + memInfo.AnonPages;
/* NUMA node memory usage */
if (resource_config[index].percent == 1) {
memUsage = (double) memInfo.AnonPages / memTotal;
memUsage = memUsage * 100;
dlog("Memory Usage %s: %.2f%% \n", resource_name.c_str(), memUsage);
} else {
memUsage = (double) avail / MiB;
dlog("Memory Available %s: %.2f MB \n", resource_name.c_str(), memUsage);
}
/* initialize a new dynamic resource for the NUMA node if it does not already exist */
save_dynamic_mem_resource ( resource_name, criticality, memUsage, resource_config[index].percent,
absolute_thresholds, MEMORY_ALARM_ID );
}
if (HUGEPAGES_NODE == 1) {
/* huge pages memory usage for the NUMA node */
if (memInfo.HugePages_Total != 0){
if (resource_config[index].percent == 1){
memUsageHuge = (double) (memInfo.HugePages_Total - memInfo.HugePages_Free) / memInfo.HugePages_Total;
memUsageHuge = memUsageHuge * 100;
dlog("Memory Usage %s: %.2f%% \n", resource_name_huge.c_str(), memUsageHuge);
} else {
memUsageHuge = (double) memInfo.HugePages_Free * (memInfo.Hugepagesize/MiB) ;
dlog("Memory Available %s: %.2f MB \n", resource_name_huge.c_str(), memUsageHuge);
}
save_dynamic_mem_resource ( resource_name_huge, criticality, memUsageHuge, resource_config[index].percent,
absolute_thresholds, MEMORY_ALARM_ID );
}
}
resource_name_huge = "processor_hugepages_";
resource_name = "processor_";
}
}
/*****************************************************************************
*
* Name : get_cpu_time
*
* Purpose : Parse per-cpu hi-resolution scheduling stats
*
*****************************************************************************/
int get_cpu_time( unsigned long long * cpu_time )
{
#define MAX_STRING_SIZE (19)
const char *sched_stat = "/proc/schedstat";
FILE * pFile;
char cpu_line[500];
unsigned long long value;
int version = 0;
int index = 0;
char cpu_time_len[50];
if ((pFile = fopen(sched_stat, "r")) == NULL){
dlog("failed to open: /proc/schedstat \n");
return (FAIL);
}
else {
/* Parse per-cpu hi-resolution scheduling stats */
while (memset(cpu_line, 0, sizeof(cpu_line)) && (fgets((char*) &cpu_line, sizeof(cpu_line), pFile) != NULL)) {
if (version != 15){
/* only version 15 is supported */
if (sscanf(cpu_line, "version %llu", &value) == 1) {
version = (int) value;
}
}
else if ((strstr(cpu_line, "cpu") != NULL) && (version == 15))
{
sscanf(cpu_line, "%*s %*s %*s %*s %*s %*s %*s %49s ",cpu_time_len);
if (((unsigned)strlen(cpu_time_len)) < MAX_STRING_SIZE) {
/* get the cpu time values for each cpu which is the 7th field */
sscanf(cpu_line, "%*s %*s %*s %*s %*s %*s %*s %llu ",&value);
cpu_time[index++] = value;
}
else {
elog("%s exceeded 2^64 for cpu stats cannot calculate cpu usage\n", cpu_time_len);
cpu_time[index++] = 0;
}
}
}
fclose(pFile);
}
return (PASS);
}
/*****************************************************************************
*
* Name : cpu_monitoring_init
*
* Purpose : Get the base cpu list if running on a compute. Also get the number
* of cpus from: /proc/cpuinfo
*****************************************************************************/
void cpu_monitoring_init()
{
string base_cpu="";
FILE * pFile;
string delimiter = ",", delimiterTwo = "-";
size_t pos = 0;
string token;
char cpu_line[100];
const char *cpu_info = "/proc/cpuinfo";
char processor[20];
pFile = fopen (COMPUTE_RESERVED_CONF , "r");
if (pFile != NULL){
ilog("File %s is present\n", COMPUTE_RESERVED_CONF);
ifstream fin( COMPUTE_RESERVED_CONF );
string line;
while( getline( fin, line ) ) {
/* process each line */
if( line.find ("PLATFORM_CPU_LIST=") != string::npos ) {
stringstream ss( line );
getline( ss, base_cpu, '=' ); // token = string before =
getline( ss, base_cpu, '=' ); // token = string after =
ilog("Found PLATFORM_CPU_LIST set to %s in file %s\n", base_cpu.c_str(), COMPUTE_RESERVED_CONF);
}
}
fclose (pFile);
}
if (base_cpu.compare("") != 0)
{
/* get base cpus if they are available */
if ((pos = base_cpu.find(delimiter)) != string::npos) {
/* if the base cpus are listed with a comma, ex: 1,2 */
base_cpu = base_cpu + delimiter;
while ((pos = base_cpu.find(delimiter)) != string::npos) {
token = base_cpu.substr(0, pos);
included_cpu[num_base_cpus++] = atoi(token.c_str());
base_cpu.erase(0, pos + delimiter.length());
}
} else if ((pos = base_cpu.find(delimiterTwo)) != string::npos) {
/* if the base cpus are listed with a dash, ex: 1-3 */
base_cpu = base_cpu + delimiterTwo;
token = base_cpu.substr(0, pos);
int first_cpu = atoi(token.c_str());
base_cpu.erase(0, pos + delimiterTwo.length());
pos = base_cpu.find(delimiterTwo);
token = base_cpu.substr(0, pos);
int last_cpu = atoi(token.c_str());
/* loop through the list of base cpus */
for (num_base_cpus=0; num_base_cpus<=(last_cpu - first_cpu); num_base_cpus++){
included_cpu[num_base_cpus++] = first_cpu++;
}
}
if (num_base_cpus == 0) {
/* only one base cpu available */
included_cpu[num_base_cpus++] = atoi(base_cpu.c_str());
}
}
ilog("Number of base CPUs for this node is %d \n", num_base_cpus);
/* get the number of cpus */
if ((pFile = fopen(cpu_info, "r")) == NULL){
wlog("failed to open: /proc/cpuinfo \n");
}
else {
/* Parse per-cpu hi-resolution scheduling stats */
while (memset(cpu_line, 0, sizeof(cpu_line)) && (fgets((char*) &cpu_line, sizeof(cpu_line), pFile) != NULL)) {
sscanf(cpu_line, "%19s %*s %*s", processor);
if (strcmp(processor, "processor") == 0) {
num_cpus++;
}
}
fclose(pFile);
}
ilog("Number of CPUs for this node is %d \n", num_cpus);
}
/*****************************************************************************
*
* Name : calculate_linux_usage
*
* Purpose : Calculate the cpu usage for Linux cards: controller, compute, storage
* The calculation runs as a delta. The first time the function is called no
* valid cpu calculation occurs. From the second time onwards, the cpu uasge is
* calculated by taking the delta from the previous time the function was called
*
*****************************************************************************/
int calculate_linux_usage( resource_config_type * ptr )
{
double delta_seconds;
unsigned long long cpu_occupancy[num_cpus];
unsigned long long cpu_delta_time;
unsigned long long total_avg_cpu = 0;
unsigned int counted_cpu=0;
int rc;
unsigned long long cpu_time[num_cpus];
if (cpu_time_initial.size() == 0) {
/* get the cpu time initially if the first cpu time does not exist */
rc = get_cpu_time( cpu_time );
/* get the first timestamp */
time(&t1);
if (rc != PASS)
{
wlog("Failed get_cpu_time \n");
return (FAIL);
}
for (int x=0; x<num_cpus; x++){
/* add the pointer value to a vector for permanent storage */
cpu_time_initial.push_back(cpu_time[x]);
dlog2("cpu_time_initial for Cpu%d set to %llu \n", x, cpu_time[x]);
}
ptr->resource_value = 0;
}
else {
/* get the later cpu time if the first cpu time exists */
rc = get_cpu_time( cpu_time );
if (rc != PASS)
{
wlog("Failed get_cpu_time \n");
return (FAIL);
}
/* get the later timestamp */
time(&t2);
for (int x=0; x<num_cpus; x++){
/* add the pointer value to a vector for permanent storage */
cpu_time_later.push_back(cpu_time[x]);
dlog2("cpu_time_later for Cpu%d %llu \n", x, cpu_time[x]);
}
delta_seconds = difftime(t2, t1);
if (num_base_cpus == 0)
{
/*this is a controller or storage node there are no vswitch or pinned cpus */
for (int j=0; j<num_cpus; j++)
{
/* calculate the cpu runtime for the specific cpu core */
dlog("Cpu%d delta_seconds: %f\n",j, delta_seconds);
dlog("Cpu%d later cycles: %llu , early cycles: %llu \n", j, cpu_time_later.at(j), cpu_time_initial.at(j));
cpu_delta_time = ((cpu_time_later.at(j) - cpu_time_initial.at(j)) / 1000000 /delta_seconds);
cpu_occupancy[j] = (100*(cpu_delta_time))/1000;
dlog("Cpu%d cpu_delta_time: %llu\n",j, cpu_delta_time);
dlog("Cpu%d cpu_occupancy: %llu\n",j, cpu_occupancy[j]);
total_avg_cpu += cpu_occupancy[j];
counted_cpu++;
}
} else {
/* this is a compute node, do not include vswitch or pinned cpus in calculation */
for (int j=0; j<num_base_cpus; j++)
{
/* calculate the cpu runtime for the specific cpu core */
dlog("Cpu%d delta_seconds: %f\n", included_cpu[j], delta_seconds);
dlog("Cpu%d later cycles: %llu , early cycles: %llu \n", included_cpu[j], cpu_time_later.at(included_cpu[j]), cpu_time_initial.at(included_cpu[j]));
cpu_delta_time = ((cpu_time_later.at(included_cpu[j]) - cpu_time_initial.at(included_cpu[j])) / 1000000 /delta_seconds);
cpu_occupancy[j] = (100*(cpu_delta_time))/1000;
dlog("Cpu%d cpu_delta_time: %llu\n", included_cpu[j], cpu_delta_time);
dlog("Cpu%d cpu_occupancy: %llu\n", included_cpu[j], cpu_occupancy[j]);
total_avg_cpu += cpu_occupancy[j];
counted_cpu++;
}
}
dlog("total_avg_cpu: %llu , counted_cpu: %d \n", total_avg_cpu, counted_cpu);
ptr->resource_value = (double) (total_avg_cpu / counted_cpu);
/* clear the old cpu times and set the current times as the old times */
cpu_time_initial.clear();
for (int x=0; x<num_cpus; x++){
/* copy the current cpu times to the earlier ones */
cpu_time_initial.push_back(cpu_time_later.at(x));
}
/* clear the current cpu times as they will be updated the next time around */
cpu_time_later.clear();
/* set the previous time as the current time */
t1 = t2;
#define LINUX_CPU_LOG_VALUE_STEP (10)
if ( log_value ( ptr->resource_value,
ptr->resource_prev,
LINUX_CPU_LOG_VALUE_STEP ))
{
plog("%s: %.2f%% (average)\n", ptr->resource, ptr->resource_value);
}
}
return (PASS);
}
/* Read the node UUID from the: /etc/platform/platform.conf file */
void _readUUID ()
{
FILE * pFile;
const char *platformFile = "/etc/platform/platform.conf";
pFile = fopen (platformFile , "r");
if (pFile != NULL) {
ifstream fin( platformFile );
string line;
while( getline( fin, line ) ) {
/* process each line */
if( line.find ("UUID=") != string::npos ) {
stringstream ss( line );
getline( ss, hostUUID, '=' ); // token = string before =
getline( ss, hostUUID, '=' ); // token = string after =
}
}
fclose (pFile);
}
}
/*****************************************************************************
*
* Name : _load_rmon_interfaces
*
* Purpose : Update the monitored network interfaces from the:
* /etc/plaform/interfaces file
*****************************************************************************/
void _load_rmon_interfaces ()
{
rmon_socket_type * sock_ptr = rmon_getSock_ptr ();
/* initialize interface monitoring */
for ( int j = 0 ; j < _rmon_ctrl_ptr->interface_resources; j++ )
{
init_physical_interfaces ( &interface_resource_config[j] );
}
for (int i=0; i<_rmon_ctrl_ptr->interface_resources; i++)
{
if ( interface_resource_config[i].interface_used == true )
{
/* set the link state for all the primary physical interfaces */
if ( get_link_state ( sock_ptr->ioctl_sock, interface_resource_config[i].interface_one, &interface_resource_config[i].link_up_and_running ) )
{
interface_resource_config[i].link_up_and_running = false ;
interface_resource_config[i].resource_value = INTERFACE_DOWN;
wlog ("Failed to query %s operational state ; defaulting to down\n", interface_resource_config[i].interface_one) ;
}
else
{
ilog ("%s link is: %s\n", interface_resource_config[i].interface_one, interface_resource_config[i].link_up_and_running ? "Up" : "Down" );
if (interface_resource_config[i].link_up_and_running)
{
interface_resource_config[i].resource_value = INTERFACE_UP;
}
else
{
interface_resource_config[i].resource_value = INTERFACE_DOWN;
interface_resource_config[i].failed = true;
}
}
if (interface_resource_config[i].lagged == true)
{
/* set the link state for all the lagged physical interfaces */
if ( get_link_state ( sock_ptr->ioctl_sock, interface_resource_config[i].interface_two, &interface_resource_config[i].link_up_and_running ) )
{
interface_resource_config[i].link_up_and_running = false ;
wlog ("Failed to query %s operational state ; defaulting to down\n", interface_resource_config[i].interface_two) ;
}
else
{
ilog ("%s link is: %s\n", interface_resource_config[i].interface_two, interface_resource_config[i].link_up_and_running ? "Up" : "Down" );
if (interface_resource_config[i].link_up_and_running)
{
interface_resource_config[i].resource_value_lagged = INTERFACE_UP;
}
else
{
interface_resource_config[i].resource_value_lagged = INTERFACE_DOWN;
interface_resource_config[i].failed = true;
}
}
}
}
}
for ( int j = 0 ; j < _rmon_ctrl_ptr->interface_resources; j++ )
{
interface_alarming_init ( &interface_resource_config[j] );
}
}
/*****************************************************************************
*
* Name : resource_stall_monitor
*
* Purpose : Detects stalls in the resource monitoring threads
******************************************************************************/
int resource_stall_monitor ( resource_config_type * ptr, pid_t tid, pid_t pid)
{
#define MAX_SCHEDSTAT_LEN (128)
char file_path [MAX_FILENAME_LEN] ;
char schedstat [MAX_SCHEDSTAT_LEN] ;
FILE * fp ;
int rc = PASS;
unsigned long long nr_switches_old = t_data.nr_switches_count;
snprintf ( &file_path[0], MAX_FILENAME_LEN, "/proc/%d/task/%d/schedstat", pid, tid );
fp = fopen (file_path, "r" );
if ( fp )
{
/* check to see if the thread is stalled */
memset ( schedstat, 0 , MAX_SCHEDSTAT_LEN );
if ( fgets ( &schedstat[0], MAX_SCHEDSTAT_LEN, fp) != NULL)
{
if ( sscanf ( schedstat, "%*s %*s %llu", &t_data.nr_switches_count) >= 1 )
{
dlog ("%s: nr_count: %llu, nr_count_old: %llu \n", ptr->resource, t_data.nr_switches_count, nr_switches_old);
if ((nr_switches_old != t_data.nr_switches_count) && (ptr->failed))
{
/* Clear the stall monitor alarm */
ilog("%s thread has unstalled \n", ptr->resource);
ptr->sev = SEVERITY_CLEARED;
t_data.nr_switches_count = 0;
resourceStageChange ( ptr, RMON_STAGE__FINISH );
}
}
else
{
wlog ("Failed to get schedstat from (%s)\n", file_path);
rc = FAIL;
}
}
else
{
wlog ("failed to read from (%s)\n", file_path );
rc = FAIL;
}
fclose(fp);
}
else
{
wlog ("Failed to open (%s)\n", file_path);
rc = FAIL;
}
if ((((nr_switches_old == t_data.nr_switches_count) && (ptr->sev != SEVERITY_MAJOR))) ||
(rc == FAIL))
{
/* thread has stalled raise alarm */
elog("%s thread has stalled \n", ptr->resource);
ptr->sev = SEVERITY_MAJOR;
ptr->failed = true;
resourceStageChange ( ptr, RMON_STAGE__MANAGE );
}
return rc;
}
/*****************************************************************************
*
* Name : check_instance_file
*
* Purpose : Thread spawned by rmon to check if: /etc/nova/instances is mounted.
* It needs to be a thread because of NFS hang issues.
*
*****************************************************************************/
void *check_instance_file(void *threadarg)
{
struct thread_data *res_data;
FILE * pFile;
FILE *testFile;
string line;
struct stat p;
const char *instances_dir = "/etc/nova/instances";
const char *test_file = "/etc/nova/instances/.rmon_test";
res_data = (struct thread_data *) threadarg;
pthread_mutex_lock(&lock);
res_data->thread_running = true;
res_data->tid = syscall(SYS_gettid);
pthread_mutex_unlock(&lock);
dlog("%s process id: %d, thread id: %d \n", res_data->resource->resource, res_data->pid, res_data->tid);
res_data->resource_usage = NOT_MOUNTED;
pFile = fopen (MOUNTS_DIR , "r");
/* query /proc/mounts and make sure the /etc/nova/instances file system is there */
if (pFile != NULL)
{
ifstream fin( MOUNTS_DIR );
while( getline( fin, line ) )
{
/* process each line */
if( line.find (instances_dir) != string::npos )
{
/* the mount is present */
res_data->resource_usage = MOUNTED;
break;
}
}
fclose (pFile);
}
if ( res_data->resource_usage == MOUNTED )
{
/* put the test file in and check that it is accessible */
testFile = fopen(test_file, "w");
if (testFile != NULL)
{
fclose (testFile);
if( remove( test_file ) != 0 )
{
elog("Failure in removing rmond test file: %s \n", test_file);
}
}
else
{
res_data->resource_usage = NOT_MOUNTED;
}
}
if (res_data->resource_usage == NOT_MOUNTED)
{
/* fail the resource */
stat (COMPUTE_CONFIG_PASS, &p);
if ((p.st_ino != 0 ) || (p.st_dev != 0))
{
pthread_mutex_lock(&lock);
if (res_data->resource->sev != SEVERITY_MAJOR)
{
res_data->resource->sev = SEVERITY_MAJOR;
res_data->resource->failed = true;
resourceStageChange ( res_data->resource, RMON_STAGE__MANAGE );
}
pthread_mutex_unlock(&lock);
}
}
else if ((res_data->resource_usage == MOUNTED) && (res_data->resource->failed))
{
pthread_mutex_lock(&lock);
res_data->resource->sev = SEVERITY_CLEARED;
resourceStageChange ( res_data->resource, RMON_STAGE__FINISH );
pthread_mutex_unlock(&lock);
}
pthread_mutex_lock(&lock);
res_data->thread_running = false;
pthread_mutex_unlock(&lock);
pthread_exit(NULL);
}
/*****************************************************************************
*
* Name : postPMs
*
* Purpose : create samples for each resource in Ceilometer
*
*****************************************************************************/
int _postPMs ()
{
char meta_data[MAX_LEN];
if ( hostUUID.empty() )
{
/* keep trying to get the host UUID if it is not present */
_readUUID();
}
if ( !hostUUID.empty() )
{
// indicate the platform hostname as metadata for all resources
char *hoststring = strdup(_rmon_ctrl_ptr->my_hostname);
if (hoststring) {
char *host = strtok(hoststring,"=");
host = strtok(NULL, "=");
snprintf(&meta_data[0], MAX_LEN, "{\"host\":\"%s\"}", host);
free(hoststring);
}
for ( int i = 0 ; i < _rmon_ctrl_ptr->resources ; i++ )
{
ostringstream strs;
strs << resource_config[i].resource_value ;
string res_val = strs.str();
if (strcmp(resource_config[i].resource, CPU_RESOURCE_NAME) == 0) {
/* cpu resource pm */
generate_ceilometer_pm ( hostUUID, "platform.cpu.util", "delta", "%",
res_val, string(meta_data) );
}
else if (strcmp(resource_config[i].resource, MEMORY_RESOURCE_NAME) == 0) {
/* memory resource pm */
if (resource_config[i].percent == 1) {
generate_ceilometer_pm ( hostUUID, "platform.mem.util", "delta", "%",
res_val, string(meta_data) );
} else {
generate_ceilometer_pm ( hostUUID, "platform.mem.util", "gauge", "MB",
res_val, string(meta_data) );
}
}
else if (strcmp(resource_config[i].resource, FS_RESOURCE_NAME) == 0) {
/* filesystem resource pm */
if (resource_config[i].percent == 1) {
generate_ceilometer_pm ( hostUUID, "platform.fs.util", "delta", "%",
res_val, string(meta_data) );
} else {
generate_ceilometer_pm ( hostUUID, "platform.fs.util", "gauge", "MB",
res_val, string(meta_data) );
}
}
} // end of resource loop
}
return (PASS);
}
/*****************************************************************************
*
* Name : _get_events
*
* Purpose : query each resource and extract the required usage values
*
*****************************************************************************/
extern bool is_cpe ( void );
extern bool is_worker ( void );
void _get_events (void)
{
int rc;
string v_cpu;
FILE * pFile;
if ( _rmon_ctrl_ptr->clients == 0 )
{
wlog ("Monitoring with no registered clients\n");
}
for ( int i = 0 ; i < _rmon_ctrl_ptr->resources ; i++ )
{
const char *resource = resource_config[i].resource;
ilog_throttled ( resource_config[i].resource_monitor_throttle, 120,
"Monitoring '%s'\n",
resource );
if (strcmp(resource, CPU_RESOURCE_NAME) == 0)
{
/* linux cards: controller, compute and storage cpu utilization */
rc = calculate_linux_usage( &resource_config[i] );
if ( rc == PASS )
{
/* get if the resource is failed to be used by resource handler */
process_failures ( &resource_config[i]);
}
}
else if (!strcmp(resource, V_CPU_RESOURCE_NAME) ||
!strcmp(resource, V_MEMORY_RESOURCE_NAME) ||
!strcmp(resource, V_PORT_RESOURCE_NAME) ||
!strcmp(resource, V_INTERFACE_RESOURCE_NAME) ||
!strcmp(resource, V_LACP_INTERFACE_RESOURCE_NAME) ||
!strcmp(resource, V_OVSDB_RESOURCE_NAME) ||
!strcmp(resource, V_NETWORK_RESOURCE_NAME) ||
!strcmp(resource, V_OPENFLOW_RESOURCE_NAME))
{
/* ensure that configuration has completed before computing
* vswitch resource utilization */
if ( !daemon_is_file_present ( CONFIG_COMPLETE_WORKER ) )
continue ;
pFile = fopen (COMPUTE_VSWITCH_DIR , "r");
if (pFile != NULL){
fclose (pFile);
}
else
{
wlog ("%s failed to open %s\n", resource, COMPUTE_VSWITCH_DIR);
}
}
else if (strstr(resource_config[i].resource, V_MEMORY_RESOURCE_NAME) != NULL)
{
/* vswitch memory with specific sockets */
/* skip these ones as they are already taken care of above */
}
else if(strcmp(resource, REMOTE_LOGGING_RESOURCE_NAME) == 0)
{
rmonHdlr_remotelogging_query(&resource_config[i]);
}
else if (strcmp(resource, INSTANCE_RESOURCE_NAME) == 0)
{
/* do not perform this check if we are not on a compute node.
* its not valid on storage not combo load */
if ( !is_worker () )
continue ;
if ( !daemon_is_file_present ( CONFIG_COMPLETE_WORKER ) )
continue ;
/* nova instances mount check */
pFile = fopen (COMPUTE_VSWITCH_DIR , "r");
if (pFile != NULL)
{
rc = PASS ;
pthread_mutex_lock(&lock);
if (!t_data.thread_running)
{
pthread_attr_t attr ;
t_data.resource = &resource_config[i];
pthread_attr_init (&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
/* launch a thread to monitor the /etc/nova/instances mount */
rc = pthread_create(&thread, &attr, check_instance_file, (void *) &t_data);
if (rc)
{
elog("%s ERROR; return code from pthread_create() is %d\n",
resource, rc);
}
pthread_attr_destroy (&attr);
}
else
{
/* If thread is still running check that it is not stalled */
resource_stall_monitor(&resource_config[i], t_data.tid, t_data.pid);
}
pthread_mutex_unlock(&lock);
fclose (pFile);
}
}
else if (strcmp(resource, MEMORY_RESOURCE_NAME) == 0) {
/* memory utilization */
calculate_memory_usage(i);
/* get if the resource is failed to be used by resource handler */
if (resource_config[i].percent == PERCENT_USED) {
process_failures ( &resource_config[i]);
} else {
process_failures_absolute ( &resource_config[i]);
}
}
else if ((strcmp(resource, V_CINDER_THINPOOL_RESOURCE_NAME) == 0) &&
(resource_config[i].alarm_status == ALARM_ON)) {
/* virtual thin pool space utilization */
rc = calculate_virtual_space_usage(i, V_CINDER_THINPOOL_RESOURCE_NAME);
/* only check resource for fail and clear if it is active */
if (rc == PASS) {
if (resource_config[i].percent == PERCENT_USED) {
/* get if the resource is failed to be used by resource handler */
process_failures (&resource_config[i]);
} else {
process_failures_absolute (&resource_config[i]);
}
}
}
else if ((strcmp(resource, V_NOVA_THINPOOL_RESOURCE_NAME) == 0) &&
(resource_config[i].alarm_status == ALARM_ON)){
/* do not perform this check if we are not on a compute node.
* its not valid on storage not combo load */
if ( !is_worker () && !is_cpe () )
continue ;
if ( !daemon_is_file_present ( CONFIG_COMPLETE_WORKER ) )
continue ;
/* virtual thin pool space utilization */
rc = calculate_virtual_space_usage(i, V_NOVA_THINPOOL_RESOURCE_NAME);
/* only check resource for fail and clear if it is active */
if (rc == PASS) {
if (resource_config[i].percent == PERCENT_USED) {
/* get if the resource is failed to be used by resource handler */
process_failures (&resource_config[i]);
} else {
process_failures_absolute (&resource_config[i]);
}
}
}
else if (strcmp(resource, FS_RESOURCE_NAME) == 0) {
/* file system utilization */
/* do nothing as we calculate individual file system location and not the total */
}
else {
/* dynamic file system resource */
pthread_mutex_lock(&lock);
if ((resource_config[i].alarm_status == ALARM_ON) && (modifyingResources == false))
{
/* only calculate the resource usage if file systems aren't being added */
calculate_fs_usage( &resource_config[i] );
/* only check resource for fail and clear if it is active */
if (resource_config[i].percent == PERCENT_USED) {
/* get if the resource is failed to be used by resource handler */
process_failures ( &resource_config[i]);
} else {
process_failures_absolute ( &resource_config[i]);
}
}
else if ((resource_config[i].alarm_status == ALARM_OFF) && (modifyingResources == false)
&& (resource_config[i].failed == true))
{
//send a clear message
send_clear_msg(i);
// we need to clear the resource's alarm if there was any set for this resource
clear_alarm_for_resource(&resource_config[i]);
}
pthread_mutex_unlock(&lock);
}
} // end of rmon resources
/*
* since interface resources are event based resourcs, i.e.
* they would only be called when netlink socket reports a
* link state event, we need to run a periodic audit on them
* as part of RMON event audit.
* This audit shall resend interface degrade statuses to maintaince
* if interface is in failed state
*/
for ( int j = 0; j < _rmon_ctrl_ptr->interface_resources; j++ )
{
if ( interface_resource_config[j].interface_used &&
interface_resource_config[j].failed == true )
{
send_interface_msg ( &interface_resource_config[j],
_rmon_ctrl_ptr->clients );
}
}
}
int kill_running_process ( int pid )
{
int result = kill ( pid, 0 );
if ( result == 0 )
{
result = kill ( pid, SIGKILL );
if ( result == 0 )
{
wlog ("NTP process kill succeeded (%d)\n", pid );
}
else
{
elog ("NTP process kill failed (%d)\n", pid );
}
}
return (PASS);
}
/* SIGCHLD handler support - for waitpid */
static bool rmon_sigchld_received = false ;
void daemon_sigchld_hdlr ( void )
{
dlog("Received SIGCHLD ...\n");
int status = 0;
pid_t tpid = 0;
while ( 0 < ( tpid = waitpid ( -1, &status, WNOHANG | WUNTRACED )))
{
dlog("NTP query script returned WIFEXITED:%d and WEXITSTATUS:%d for pid:%d\n", WIFEXITED(status), WEXITSTATUS(status), tpid);
if (tpid == ntp_child_pid)
{
rmon_sigchld_received = true ;
/* no need to wait for a timeout since we got a response, force a ring */
rmonTimer_ntp.ring = true;
ntp_status = WEXITSTATUS(status);
}
else
{
dlog ("PID:%d lookup failed ; reaped likely after timeout\n", tpid );
ntp_status = NTP_ERROR;
}
}
}
int ntp_audit_handler ( )
{
if ( ntp_stage >= NTP_STAGE__STAGES )
{
wlog ("Invalid ntp_stage (%d) ; correcting\n", ntp_stage );
ntpStageChange ( NTP_STAGE__BEGIN);
}
switch ( ntp_stage )
{
// First state
case NTP_STAGE__BEGIN:
{
mtcTimer_start ( rmonTimer_ntp, rmon_timer_handler, _rmon_ctrl_ptr->ntp_audit_period );
dlog ("Start NTP period timer (%d secs) %p\n", _rmon_ctrl_ptr->ntp_audit_period, rmonTimer_ntp.tid);
ntpStageChange ( NTP_STAGE__EXECUTE_NTPQ );
break ;
}
// Execute the ntpq command
case NTP_STAGE__EXECUTE_NTPQ:
{
if ( rmonTimer_ntp.ring == true ) //wake up from NTP period
{
ntp_status = PASS;
mtcTimer_start ( rmonTimer_ntp, rmon_timer_handler, _rmon_ctrl_ptr->ntpq_cmd_timeout );
dlog ("Start NTPQ command timer (%d secs) %p\n", _rmon_ctrl_ptr->ntpq_cmd_timeout, rmonTimer_ntp.tid);
// Execute the ntpq command
int rc = query_ntp_servers();
if (rc != PASS)
{
elog ("NTP execute_status_command returned a failure (%d)\n", rc);
ntp_status = NTP_ERROR;
}
ntpStageChange ( NTP_STAGE__EXECUTE_NTPQ_WAIT );
}
break ;
}
// Wait for the ntpq command to finish and process results
case NTP_STAGE__EXECUTE_NTPQ_WAIT:
{
// Give the command time to execute. The daemon_sigchld_hdlr will force
// a ring when the command execute successfully or returns a failure
if ( ( rmonTimer_ntp.ring == true) || (ntp_status == NTP_ERROR ) )
{
// Stop the NTP timer if still running
if ( rmonTimer_ntp.tid )
{
mtcTimer_stop ( rmonTimer_ntp );
}
if (( !rmon_sigchld_received) || (ntp_status == NTP_ERROR))
{
if ( rmon_sigchld_received == false )
{
elog ("NTPQ command execution timed out (pid:%d)\n", ntp_child_pid );
}
elog ("NTPQ returned an execution failure (rc:%d) (pid:%d)\n", ntp_status, ntp_child_pid);
if (ntp_child_pid != 0)
{
kill_running_process ( ntp_child_pid );
}
}
else
{
dlog ("NTPQ command was successful ; analyzing results\n");
ntp_query_results(ntp_status);
}
ntpStageChange ( NTP_STAGE__BEGIN );
ntp_child_pid = 0;
rmon_sigchld_received = false;
}
break;
}
default:
{
elog ("NTP invalid ntp_stage (%d)\n", ntp_stage );
/* Default to first state for invalid case. there is an issue then it will be detected */
ntpStageChange ( NTP_STAGE__BEGIN );
}
}
return (PASS);
}
/*****************************************************************************
*
* Name : rmon_service
*
* Purpose : main loop for monitoring resources
*
*****************************************************************************/
void rmon_service (rmon_ctrl_type * ctrl_ptr)
{
fd_set readfds;
struct timeval waitd;
std::list<int> socks;
rmon_socket_type * sock_ptr = rmon_getSock_ptr ();
/* initialize FM handler */
rmon_fm_init();
/* ignore SIGPIPE on swacts */
signal(SIGPIPE, SIG_IGN);
/* initialize the memory accounting: either Strict or OOM */
init_memory_accounting();
/* initialize the cpu monitoring defaults */
cpu_monitoring_init();
_readUUID();
/* Start an event timer for the interval of the resources being monitored */
ilog ("Starting 'Event Monitor' timer (%d secs) \n", ctrl_ptr->audit_period);
mtcTimer_start ( rmonTimer_event, rmon_timer_handler, 1 );
ilog ("Starting 'PM Monitor' timer (%d secs) \n", ctrl_ptr->pm_period);
mtcTimer_start ( rmonTimer_pm, rmon_timer_handler,ctrl_ptr->pm_period);
if (is_controller())
{
ntp_stage = NTP_STAGE__BEGIN;
}
/* Get an Authentication Token */
ilog ("%s Requesting initial token\n", ctrl_ptr->my_hostname );
tokenEvent.status = tokenUtil_new_token ( tokenEvent, ctrl_ptr->my_hostname );
if ( tokenEvent.status != PASS )
{
elog ("Failed to get authentication token (%d)\n", tokenEvent.status);
if ( tokenEvent.base )
{
slog ("%s token base:%p\n",
ctrl_ptr->my_hostname,
tokenEvent.base);
}
}
/* service all the register and deregister requests in the queue */
rmon_alive_notification( _rmon_ctrl_ptr->clients );
ilog ("registered clients: %d\n", _rmon_ctrl_ptr->clients);
#ifdef WANT_FS_MONITORING
/* Initialize the resource specific configuration */
for (int j=0; j<_rmon_ctrl_ptr->resources; j++)
{
if ( strcmp(resource_config[j].resource, FS_RESOURCE_NAME) == 0 ) {
/* determine whether percent or absolute values are used */
/* determine if virtual thin pool memory usage alarm should be on or off */
fs_percent = resource_config[j].percent;
}
}
/* add the static filesystem resources */
process_static_fs_file();
/* initialize the resource alarms */
for (int j=0; j<_rmon_ctrl_ptr->resources; j++)
{
rmon_alarming_init ( &resource_config[j] );
}
/* add any dynamic resources from before */
add_dynamic_fs_resource(false);
#else
ilog("static filesystem monitoring moved to collectd\n");
#endif
/* Clear any stale dynamic alarms that can be caused by dynamic resources. */
/* An alarm become stale for example if it was raised against a local volumn group (lvg) and */
/* later on the lvg is deleted. The node will come up and the lvg resource will not longer exist and */
/* it's related alarms not refreshed. Dynamic alarms are any alarms which it's resource can be */
/* provisioned. */
AlarmFilter alarmFilter;
unsigned int max_alarms=75;
char alarm_to_search[FM_MAX_BUFFER_LENGTH];
fm_alarm_id alarm_id;
snprintf(alarm_id, FM_MAX_BUFFER_LENGTH, FS_ALARM_ID);
SFmAlarmDataT *active_alarms = (SFmAlarmDataT*) calloc (max_alarms, sizeof (SFmAlarmDataT));
if (active_alarms != NULL)
{
/* get all the current alarms with id of FS_ALARM_ID which are alarms related to the file system */
/* fm_get_faults_by_id returns the number of alarms found */
if (fm_get_faults_by_id( &alarm_id, active_alarms, &max_alarms) == FM_ERR_OK)
{
bool found = false;
for ( unsigned int i = 0; i < max_alarms; i++ )
{
/* only get the 100.104 alarms */
if ((strncmp((active_alarms+i)->alarm_id, FS_ALARM_ID, sizeof((active_alarms+i)->alarm_id)) == 0)
&& (strstr((active_alarms+i)->entity_instance_id, _rmon_ctrl_ptr->my_hostname) != NULL) )
{
found = false;
for (int j=0; j<_rmon_ctrl_ptr->resources; j++)
{
/* since we build the entity_instance_id with multiple data we must recreate it */
snprintf(alarm_to_search, FM_MAX_BUFFER_LENGTH, "%s.volumegroup=%s", _rmon_ctrl_ptr->my_hostname, resource_config[j].resource);
if (strncmp(alarm_to_search, (active_alarms+i)->entity_instance_id, sizeof(alarm_to_search)) == 0)
{
found = true;
break;
}
snprintf(alarm_to_search, FM_MAX_BUFFER_LENGTH, "%s.filesystem=%s", _rmon_ctrl_ptr->my_hostname, resource_config[j].resource);
if (strncmp(alarm_to_search, (active_alarms+i)->entity_instance_id, sizeof(alarm_to_search)) == 0)
{
found = true;
break;
}
// We found the resource but lets check if the alarm is enable for it, if it's not
// we want to clear that alarm
if (found)
{
if (resource_config[j].alarm_status == ALARM_OFF)
{
found = false;
}
}
}
if (!found)
{
/* the alarm did not match any current resources so let's clear it */
snprintf(alarmFilter.alarm_id, FM_MAX_BUFFER_LENGTH, (active_alarms+i)->alarm_id );
snprintf(alarmFilter.entity_instance_id, FM_MAX_BUFFER_LENGTH, (active_alarms+i)->entity_instance_id);
ilog ("Clearing stale alarm %s for entity instance id: %s", (active_alarms+i)->alarm_id, (active_alarms+i)->entity_instance_id);
if (rmon_fm_clear(&alarmFilter) != FM_ERR_OK)
{
wlog ("Failed to clear stale alarm for entity instance id: %s", (active_alarms+i)->entity_instance_id);
}
}
}
}
}
free(active_alarms);
}
else
{
elog ("Failed to allocate memory for clearing stale dynamic alarms");
}
if (( sock_ptr->ioctl_sock = open_ioctl_socket ( )) <= 0 )
{
elog ("Failed to create ioctl socket");
}
/* Not monitoring address changes RTMGRP_IPV4_IFADDR | RTMGRP_IPV6_IFADDR */
if (( sock_ptr->netlink_sock = open_netlink_socket ( RTMGRP_LINK )) <= 0 )
{
elog ("Failed to create netlink listener socket");
}
/* load the current interfaces for monitoring */
_load_rmon_interfaces();
socks.clear();
socks.push_front (sock_ptr->rmon_tx_sock);
socks.push_front (sock_ptr->netlink_sock);
socks.sort();
for (;;) {
/* Accomodate for hup reconfig */
FD_ZERO(&readfds);
FD_SET(sock_ptr->rmon_tx_sock, &readfds);
FD_SET(sock_ptr->netlink_sock, &readfds);
waitd.tv_sec = 0;
waitd.tv_usec = SOCKET_WAIT ;
tokenUtil_log_refresh ();
/* This is used as a delay up to select timeout ; SOCKET_WAIT */
select( socks.back()+1, &readfds, NULL, NULL, &waitd);
if (FD_ISSET(sock_ptr->rmon_tx_sock, &readfds))
{
_rmon_ctrl_ptr->clients = rmon_service_inbox ( _rmon_ctrl_ptr->clients );
}
else if (FD_ISSET(sock_ptr->netlink_sock, &readfds))
{
dlog ("netlink socket fired\n");
if ( service_interface_events ( sock_ptr->netlink_sock, sock_ptr->ioctl_sock ) != PASS )
{
elog ("service_interface_events failed \n");
}
}
/* Manage the health of the resources */
if ( rmonTimer_event.ring == true )
{
// restart the audit period timer
mtcTimer_start ( rmonTimer_event, rmon_timer_handler, ctrl_ptr->audit_period );
/* service all the register and deregister requests in the queue */
rmon_alive_notification( _rmon_ctrl_ptr->clients );
_get_events ( );
}
if ( rmonTimer_pm.ring == true )
{
mtcTimer_start ( rmonTimer_pm, rmon_timer_handler, ctrl_ptr->pm_period );
tokenUtil_token_refresh ( tokenEvent, ctrl_ptr->my_hostname );
_postPMs();
}
/* loop through all the resource timers waiting for a ring */
for ( int j = 0 ; j < ctrl_ptr->resources ; j++ )
{
if (resource_config[j].failed == true) {
/* Run the FSM for this failed resource */
resource_handler ( &resource_config[j]);
}
}
/* loop through all the interface resources */
for ( int j = 0 ; j < ctrl_ptr->interface_resources ; j++ )
{
if (interface_resource_config[j].failed == true) {
/* Run the FSM for this failed interface */
interface_handler ( &interface_resource_config[j] );
}
}
/* loop thorough all the LVM thinpool metadata resources waiting for a ring */
for ( int j = 0; j < ctrl_ptr->thinmeta_resources; j++ )
{
if (thinmeta_resource_config[j].critical_threshold) {
// a threshold of 0 disables monitoring
if (thinmetatimer[j].ring == true) {
// restart the audit period timer
mtcTimer_start ( thinmetatimer[j], rmon_timer_handler,
thinmeta_resource_config[j].audit_period );
dlog("%s/%s running audit (resource index: %i)",
thinmeta_resource_config[j].vg_name,
thinmeta_resource_config[j].thinpool_name, j)
/* Handle resource */
int k;
for (k = THINMETA_FSM_RETRY; k > 0; k--) {
// call again the FSM in case it instructs us to RETRY
if(thinmeta_handler(&thinmeta_resource_config[j]) != RETRY) {
break;
}
}
if (k == 0) {
dlog("%s/%s too many state changes in FSM at: %i stage!",
thinmeta_resource_config[j].vg_name,
thinmeta_resource_config[j].thinpool_name,
thinmeta_resource_config[j].stage);
}
}
}
}
/* handle RMON FM interface */
rmon_fm_handler ();
daemon_signal_hdlr ();
}
}
/****************************************************************************
*
* Name : log_value
*
* Purpose : Log resource state values while avoiding log flodding for
* trivial fluxuations.
*
* Description: Recommends whether the current resource state value should
* be logged based on current, previous and step values.
*
* Caller should not generate such log if a false is returned.
*
* A true is returned if the currrent and previous resource values differ
* by +/- step amount.
*
* The caller specifies the step that can be overridden by a smaller value
* in rmond.conf:log_step value.
*
* If step is zero then a true is always returned in support of a debug mode
* where we get the current reading as a log on every audit.
*
* The callers previous value is updated to current whenever true is returned.
*
****************************************************************************/
bool log_value ( double & current, double & previous, int step )
{
/* Support step override for debug purposes
* Allows for more frequent logging */
int _step = daemon_get_cfg_ptr()->log_step ;
/* a lower value from the conf file takes precidence */
if ( _step > step )
_step = step ;
if (( round(current) >= ( round(previous) + _step )) ||
( round(current) <= ( round(previous) - _step )))
{
previous = current ;
return true ;
}
return false ;
}
View Git Blame Copy Permalink