swift/bin/swift-drive-audit
Lorcan cb20763893 Add new features to swift-drive-audit
This patch adds two new features to swift-drive-audit. The first
is an option in the drive-audit.conf file that allows the operator
to prevent the drives ever being unmounted automatically,
regardless of the amount of errors present. This could be of
benefit in very small systems consisting of only one or two drives
where the operator would like to manually unmount/fix the
particular drive(s) and minimise any potential downtime.

The second is another option in drive-audit.conf that allows the
operator to select a recon directory. This directory will then
have a drive.recon file which will keep an up-to-date record of
the swift drives and any errors associated with them. An example
of the output would be as follows:

{"/srv/node/disk2": "0", "/srv/node/disk3": "25", "/srv/node/disk0": "0",
"/srv/node/disk1": "0", "/srv/node/disk10": "0", "/srv/node/disk7": "0",
"/srv/node/disk4": "137", "/srv/node/disk5": "0", "/srv/node/disk8": "0",
"/srv/node/disk9": "0", "/srv/node/disk6": "0", "/srv/node/disk11": "60"}

This would allow the operator to monitor the errors on the swift
drives without having to spend time searching through logs. Also, if
this is accepted, it should be possible to add an option to
swift-recon that would keep track of this at a system level.

Change-Id: Ib5dacf8622b7363e070c274c7c30c8ead448a055
2014-09-25 16:27:25 +01:00

202 lines
7.6 KiB
Python
Executable File

#!/usr/bin/env python
# Copyright (c) 2010-2012 OpenStack Foundation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import datetime
import glob
import os
import re
import subprocess
import sys
from ConfigParser import ConfigParser
from swift.common.utils import backward, get_logger, dump_recon_cache, \
config_true_value
def get_devices(device_dir, logger):
devices = []
for line in open('/proc/mounts').readlines():
data = line.strip().split()
block_device = data[0]
mount_point = data[1]
if mount_point.startswith(device_dir):
device = {}
device['mount_point'] = mount_point
device['block_device'] = block_device
try:
device_num = os.stat(block_device).st_rdev
except OSError:
# If we can't stat the device, then something weird is going on
logger.error("Error: Could not stat %s!" %
block_device)
continue
device['major'] = str(os.major(device_num))
device['minor'] = str(os.minor(device_num))
devices.append(device)
for line in open('/proc/partitions').readlines()[2:]:
major, minor, blocks, kernel_device = line.strip().split()
device = [d for d in devices
if d['major'] == major and d['minor'] == minor]
if device:
device[0]['kernel_device'] = kernel_device
return devices
def get_errors(error_re, log_file_pattern, minutes):
# Assuming log rotation is being used, we need to examine
# recently rotated files in case the rotation occurred
# just before the script is being run - the data we are
# looking for may have rotated.
#
# The globbing used before would not work with all out-of-box
# distro setup for logrotate and syslog therefore moving this
# to the config where one can set it with the desired
# globbing pattern.
log_files = [f for f in glob.glob(log_file_pattern)]
try:
log_files.sort(key=lambda f: os.stat(f).st_mtime, reverse=True)
except (IOError, OSError) as exc:
logger.error(exc)
print(exc)
sys.exit(1)
now_time = datetime.datetime.now()
end_time = now_time - datetime.timedelta(minutes=minutes)
# kern.log does not contain the year so we need to keep
# track of the year and month in case the year recently
# ticked over
year = now_time.year
prev_entry_month = now_time.month
errors = {}
reached_old_logs = False
for path in log_files:
try:
f = open(path)
except IOError:
logger.error("Error: Unable to open " + path)
print("Unable to open " + path)
sys.exit(1)
for line in backward(f):
if '[ 0.000000]' in line \
or 'KERNEL supported cpus:' in line \
or 'BIOS-provided physical RAM map:' in line:
# Ignore anything before the last boot
reached_old_logs = True
break
# Solves the problem with year change - kern.log does not
# keep track of the year.
log_time_entry = line.split()[:3]
if log_time_entry[0] == 'Dec' and prev_entry_month == 'Jan':
year -= 1
prev_entry_month = log_time_entry[0]
log_time_string = '%s %s' % (year, ' '.join(log_time_entry))
try:
log_time = datetime.datetime.strptime(
log_time_string, '%Y %b %d %H:%M:%S')
except ValueError:
continue
if log_time > end_time:
for err in error_re:
for device in err.findall(line):
errors[device] = errors.get(device, 0) + 1
else:
reached_old_logs = True
break
if reached_old_logs:
break
return errors
def comment_fstab(mount_point):
with open('/etc/fstab', 'r') as fstab:
with open('/etc/fstab.new', 'w') as new_fstab:
for line in fstab:
parts = line.split()
if len(parts) > 2 and line.split()[1] == mount_point:
new_fstab.write('#' + line)
else:
new_fstab.write(line)
os.rename('/etc/fstab.new', '/etc/fstab')
if __name__ == '__main__':
c = ConfigParser()
try:
conf_path = sys.argv[1]
except Exception:
print "Usage: %s CONF_FILE" % sys.argv[0].split('/')[-1]
sys.exit(1)
if not c.read(conf_path):
print "Unable to read config file %s" % conf_path
sys.exit(1)
conf = dict(c.items('drive-audit'))
device_dir = conf.get('device_dir', '/srv/node')
minutes = int(conf.get('minutes', 60))
error_limit = int(conf.get('error_limit', 1))
recon_cache_path = conf.get('recon_cache_path', "/var/cache/swift")
log_file_pattern = conf.get('log_file_pattern',
'/var/log/kern.*[!.][!g][!z]')
error_re = []
for conf_key in conf:
if conf_key.startswith('regex_pattern_'):
error_pattern = conf[conf_key]
try:
r = re.compile(error_pattern)
except re.error:
sys.exit('Error: unable to compile regex pattern "%s"' %
error_pattern)
error_re.append(r)
if not error_re:
error_re = [
re.compile(r'\berror\b.*\b(sd[a-z]{1,2}\d?)\b'),
re.compile(r'\b(sd[a-z]{1,2}\d?)\b.*\berror\b'),
]
conf['log_name'] = conf.get('log_name', 'drive-audit')
logger = get_logger(conf, log_route='drive-audit')
devices = get_devices(device_dir, logger)
logger.debug("Devices found: %s" % str(devices))
if not devices:
logger.error("Error: No devices found!")
recon_errors = {}
for device in devices:
recon_errors[device['mount_point']] = 0
errors = get_errors(error_re, log_file_pattern, minutes)
logger.debug("Errors found: %s" % str(errors))
unmounts = 0
for kernel_device, count in errors.items():
if count >= error_limit:
device = \
[d for d in devices if d['kernel_device'] == kernel_device]
if device:
mount_point = device[0]['mount_point']
if mount_point.startswith(device_dir):
if config_true_value(conf.get('unmount_failed_device',
True)):
logger.info("Unmounting %s with %d errors" %
(mount_point, count))
subprocess.call(['umount', '-fl', mount_point])
logger.info("Commenting out %s from /etc/fstab" %
(mount_point))
comment_fstab(mount_point)
unmounts += 1
recon_errors[mount_point] = count
recon_file = recon_cache_path + "/drive.recon"
dump_recon_cache(recon_errors, recon_file, logger)
if unmounts == 0:
logger.info("No drives were unmounted")