tripleo-common/tripleo_common/utils/config.py
James Slagle 8532a397de Warn when looking up grandparent resource name
When running config-download if the deployment name property was not set
on a deployment resource from a ResourceGroup, it will trigger the code
path to lookup the grandparent resource name.

This codepath can be very slow depending on the number of nodes in the
stack.

This patch adds a warning when this is happening so that it's clear that
the name property needs to be set on the deployment resource.

Change-Id: I2cf7acef0999e1c36d4baa8c793bdaa78bd441ad
2019-09-17 12:43:58 -04:00

550 lines
25 KiB
Python

# Copyright 2016 Red Hat, Inc.
# All Rights Reserved.
#
# 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 json
import logging
import os
import re
import shutil
import six
import warnings
import yaml
import jinja2
from tripleo_common import constants
from tripleo_common.utils.safe_import import git
warnings.filterwarnings('once')
class Config(object):
def __init__(self, orchestration_client):
self.log = logging.getLogger(__name__ + ".Config")
self.client = orchestration_client
self.stack_outputs = {}
def get_server_names(self):
servers = {}
role_node_id_map = self.stack_outputs.get('ServerIdData', {})
role_net_hostname_map = self.stack_outputs.get(
'RoleNetHostnameMap', {})
for role, hostnames in role_net_hostname_map.items():
if hostnames:
names = hostnames.get(constants.HOST_NETWORK) or []
shortnames = [n.split(".%s." % constants.HOST_NETWORK)[0]
for n in names]
for idx, name in enumerate(shortnames):
if 'server_ids' in role_node_id_map:
server_id = role_node_id_map['server_ids'][role][idx]
servers[server_id] = name.lower()
return servers
def get_network_config_data(self, stack):
return self.stack_outputs.get("HostnameNetworkConfigMap")
def get_deployment_data(self, stack,
nested_depth=constants.NESTED_DEPTH):
deployments = self.client.resources.list(
stack,
nested_depth=nested_depth,
filters=dict(name=constants.TRIPLEO_DEPLOYMENT_RESOURCE),
with_detail=True)
# Sort by creation time
deployments = sorted(deployments, key=lambda d: d.creation_time)
return deployments
def get_role_from_server_id(self, stack, server_id):
server_id_data = self.stack_outputs.get('ServerIdData', {}
).get('server_ids', {})
for k, v in server_id_data.items():
if server_id in v:
return k
def get_deployment_resource_id(self, deployment):
if '/' in deployment.attributes['value']['deployment']:
deployment_stack_id = \
deployment.attributes['value']['deployment'].split('/')[-1]
deployment_resource_id = self.client.resources.get(
deployment_stack_id,
'TripleOSoftwareDeployment').physical_resource_id
else:
deployment_resource_id = \
deployment.attributes['value']['deployment']
return deployment_resource_id
def get_config_dict(self, deployment_resource_id):
deployment_rsrc = self.client.software_deployments.get(
deployment_resource_id)
config = self.client.software_configs.get(
deployment_rsrc.config_id)
return config.to_dict()
def get_jinja_env(self, tmp_path):
templates_path = os.path.join(
os.path.dirname(__file__), '..', 'templates')
self._mkdir(os.path.join(tmp_path, 'templates'))
env = jinja2.Environment(
loader=jinja2.FileSystemLoader(templates_path))
env.trim_blocks = True
return env, templates_path
def get_role_config(self):
role_config = self.stack_outputs.get('RoleConfig', {})
# RoleConfig can exist as a stack output but have a value of None
return role_config or {}
@staticmethod
def _open_file(path):
return os.fdopen(
os.open(path, os.O_WRONLY | os.O_CREAT | os.O_TRUNC, 0o600), 'w')
def _write_tasks_per_step(self, tasks, role, filepath, step):
def step_in_task(task, step):
whenexpr = task.get('when', None)
if whenexpr is None:
# If no step is defined, it will be executed for all
# steps.
return True
if not isinstance(whenexpr, list):
whenexpr = [whenexpr]
for w in whenexpr:
# make \|int optional incase forgotten; use only step digit:
match = re.search(r'step(\|int)? == ([0-9]+)$', "%s" % w)
if match:
if match.group(2) == str(step):
return True
else:
return False
return True
tasks_per_step = [task for task in tasks if step_in_task(task, step)]
with self._open_file(filepath) as conf_file:
yaml.safe_dump(tasks_per_step, conf_file, default_flow_style=False)
return tasks_per_step
def initialize_git_repo(self, dirname):
repo = git.Repo.init(dirname)
gitignore_path = os.path.join(dirname, '.gitignore')
# Ignore tarballs, which we use for the export process
if not os.path.exists(gitignore_path):
with open(gitignore_path, 'w') as f:
f.write('*.tar.gz\n')
# For some reason using repo.index.add is not working, so go
# directly to the GitCmd interface.
repo.git.add('.gitignore')
return repo
def snapshot_config_dir(self, repo, commit_message):
if repo.is_dirty(untracked_files=True):
self.log.info('Snapshotting {}'.format(repo.working_dir))
# Use repo.git.add directly as repo.index.add defaults to forcing
# commit of ignored files, which we don't want.
repo.git.add('.')
commit = repo.index.commit(commit_message)
self.log.info('Created commit {}'.format(commit.hexsha))
else:
self.log.info('No changes to commit')
def _mkdir(self, dirname):
if not os.path.exists(dirname):
try:
os.makedirs(dirname, 0o700)
except OSError as e:
message = 'Failed to create: %s, error: %s' % (dirname,
str(e))
raise OSError(message)
def create_config_dir(self, config_dir, preserve_config_dir=True):
# Create config directory
if os.path.exists(config_dir) and preserve_config_dir is False:
try:
self.log.info("Directory %s already exists, removing"
% config_dir)
shutil.rmtree(config_dir)
except OSError as e:
message = 'Failed to remove: %s, error: %s' % (config_dir,
str(e))
raise OSError(message)
def fetch_config(self, name):
# Get the stack object
stack = self.client.stacks.get(name)
self.stack_outputs = {i['output_key']: i['output_value']
for i in stack.outputs}
return stack
def validate_config(self, template_data, yaml_file):
try:
yaml.safe_load(template_data)
except (yaml.scanner.ScannerError, yaml.YAMLError) as e:
self.log.error("Config for file {} contains invalid yaml, got "
"error {}".format(yaml_file, e))
raise e
def render_network_config(self, stack, config_dir, server_roles):
network_config = self.get_network_config_data(stack)
for server, config in network_config.items():
server_deployment_dir = os.path.join(
config_dir, server_roles[server], server)
self._mkdir(server_deployment_dir)
network_config_path = os.path.join(
server_deployment_dir, "NetworkConfig")
network_config_role_path = os.path.join(
config_dir, server_roles[server], "NetworkConfig")
s_config = self.client.software_configs.get(config)
if getattr(s_config, 'config', ''):
with self._open_file(network_config_path) as f:
f.write(s_config.config)
with self._open_file(network_config_role_path) as f:
f.write(s_config.config)
def write_config(self, stack, name, config_dir, config_type=None):
# Get role data:
role_data = self.stack_outputs.get('RoleData', {})
role_group_vars = self.stack_outputs.get('RoleGroupVars', {})
role_host_vars = self.stack_outputs.get('AnsibleHostVarsMap', {})
for role_name, role in six.iteritems(role_data):
role_path = os.path.join(config_dir, role_name)
self._mkdir(role_path)
for config in config_type or role.keys():
if config in constants.EXTERNAL_TASKS:
# external tasks are collected globally, not per-role
continue
elif config == 'step_config':
filepath = os.path.join(role_path, 'step_config.pp')
with self._open_file(filepath) as step_config:
step_config.write(role[config])
elif config == 'param_config':
filepath = os.path.join(role_path, 'param_config.json')
with self._open_file(filepath) as param_config:
param_config.write(json.dumps(role[config]))
elif config == 'ansible_group_vars':
# NOTE(aschultz): ansible group vars are for specific role
# services so we merge them in with the others so they
# end up in the role vars. This means the last var
# definition wins and will override them all.
if role_name not in role_group_vars:
role_group_vars[role_name] = {}
role_group_vars[role_name].update(role[config])
else:
# NOTE(jfrancoa): Move this upgrade_tasks condition to the
# upper level once THT is adapted. We include it here to
# allow the CI to pass until THT changed is not merged.
if config == 'upgrade_tasks':
for i in range(constants.UPGRADE_STEPS_MAX):
filepath = os.path.join(role_path, '%s_step%s.yaml'
% (config, i))
self._write_tasks_per_step(role[config], role_name,
filepath, i)
try:
data = role[config]
except KeyError as e:
message = 'Invalid key: %s, error: %s' % (config,
str(e))
raise KeyError(message)
filepath = os.path.join(role_path, '%s.yaml' % config)
with self._open_file(filepath) as conf_file:
yaml.safe_dump(data,
conf_file,
default_flow_style=False)
role_config = self.get_role_config()
for config_name, config in six.iteritems(role_config):
conf_path = os.path.join(config_dir, config_name)
# Add .yaml extension only if there's no extension already
if '.' not in conf_path:
conf_path = conf_path + ".yaml"
with self._open_file(conf_path) as conf_file:
if isinstance(config, list) or isinstance(config, dict):
yaml.safe_dump(config, conf_file, default_flow_style=False)
else:
conf_file.write(config)
# Get deployment data
self.log.info("Getting deployment data from Heat...")
deployments_data = self.get_deployment_data(name)
# server_deployments is a dict of server name to a list of deployments
# (dicts) associated with that server
server_deployments = {}
# server_names is a dict of server id to server_name for easier lookup
server_names = self.get_server_names()
server_ids = dict([(v, k) for (k, v) in server_names.items()])
# server_deployment_names is a dict of server names to deployment names
# for that role. The deployment names are further separated in their
# own dict with keys of pre_deployment/post_deployment.
server_deployment_names = {}
# server_roles is a dict of server name to server role for easier
# lookup
server_roles = {}
for deployment in deployments_data:
# Check if the deployment value is the resource name. If that's the
# case, Heat did not create a physical_resource_id for this
# deployment since it does not trigger on this stack action. Such
# as a deployment that only triggers on DELETE, but this is a stack
# create. If that's the case, just skip this deployment, otherwise
# it will result in a Not found error if we try and query the
# deployment API for this deployment.
dep_value_resource_name = deployment.attributes[
'value'].get('deployment') == 'TripleOSoftwareDeployment'
# if not check the physical_resource_id
if not dep_value_resource_name:
deployment_resource_id = self.get_deployment_resource_id(
deployment)
if dep_value_resource_name or not deployment_resource_id:
warnings.warn('Skipping deployment %s because it has no '
'valid uuid (physical_resource_id) '
'associated.' %
deployment.physical_resource_id)
continue
server_id = deployment.attributes['value']['server']
config_dict = self.get_config_dict(deployment_resource_id)
# deployment_name should be set via the name property on the
# Deployment resources in the templates, however, if it's None
# or empty string, default to the name of the parent_resource.
deployment_name = deployment.attributes['value'].get(
'name') or deployment.parent_resource
if not deployment_name:
message = "The deployment name cannot be determined. It " \
"should be set via the name property on the " \
"Deployment resources in the templates."
raise ValueError(message)
try:
int(deployment_name)
except ValueError:
pass
else:
# We can't have an integer here, let's figure out the
# grandparent resource name
deployment_ref = deployment.attributes['value']['deployment']
warnings.warn('Determining grandparent resource name for '
'deployment %s. Ensure the name property is '
'set on the deployment resource in the '
'templates.' % deployment_ref)
if '/' in deployment_ref:
deployment_stack_id = deployment_ref.split('/')[-1]
else:
for link in deployment.links:
if link['rel'] == 'stack':
deployment_stack_id = link['href'].split('/')[-1]
break
else:
raise ValueError("Couldn't not find parent stack")
deployment_stack = self.client.stacks.get(
deployment_stack_id, resolve_outputs=False)
parent_stack = deployment_stack.parent
grandparent_stack = self.client.stacks.get(
parent_stack, resolve_outputs=False).parent
resources = self.client.resources.list(
grandparent_stack,
filters=dict(physical_resource_id=parent_stack))
if not resources:
message = "The deployment resource grandparent name" \
"could not be determined."
raise ValueError(message)
deployment_name = resources[0].resource_name
config_dict['deployment_name'] = deployment_name
# reset deploy_server_id to the actual server_id since we have to
# use a dummy server resource to create the deployment in the
# templates
deploy_server_id_input = \
[i for i in config_dict['inputs']
if i['name'] == 'deploy_server_id'].pop()
deploy_server_id_input['value'] = server_id
# We don't want to fail if server_id can't be found, as it's
# most probably due to blacklisted nodes. However we fail for
# other errors.
try:
server_deployments.setdefault(
server_names[server_id],
[]).append(config_dict)
except KeyError:
self.log.warning('Server with id %s is ignored from config '
'(may be blacklisted)' % server_id)
# continue the loop as this server_id is probably excluded
continue
except Exception as err:
err_msg = ('Error retrieving server name from this server_id: '
'%s with this error: %s' % server_id, err)
raise Exception(err_msg)
role = self.get_role_from_server_id(stack, server_id)
server_pre_deployments = server_deployment_names.setdefault(
server_names[server_id], {}).setdefault(
'pre_deployments', [])
server_post_deployments = server_deployment_names.setdefault(
server_names[server_id], {}).setdefault(
'post_deployments', [])
server_roles[server_names[server_id]] = role
# special handling of deployments that are run post the deploy
# steps. We have to look these up based on the
# physical_resource_id, but these names should be consistent since
# they are consistent interfaces in our templates.
if 'ExtraConfigPost' in deployment.physical_resource_id or \
'PostConfig' in deployment.physical_resource_id:
if deployment_name not in server_post_deployments:
server_post_deployments.append(deployment_name)
else:
if deployment_name not in server_pre_deployments:
server_pre_deployments.append(deployment_name)
# Make sure server_roles is populated b/c it won't be if there are no
# server deployments.
for name, server_id in server_ids.items():
server_roles.setdefault(
name,
self.get_role_from_server_id(stack, server_id))
env, templates_path = self.get_jinja_env(config_dir)
templates_dest = os.path.join(config_dir, 'templates')
self._mkdir(templates_dest)
shutil.copyfile(os.path.join(templates_path, 'heat-config.j2'),
os.path.join(templates_dest, 'heat-config.j2'))
group_vars_dir = os.path.join(config_dir, 'group_vars')
self._mkdir(group_vars_dir)
host_vars_dir = os.path.join(config_dir, 'host_vars')
self._mkdir(host_vars_dir)
for server, deployments in server_deployments.items():
deployment_template = env.get_template('deployment.j2')
for d in deployments:
server_deployment_dir = os.path.join(
config_dir, server_roles[server], server)
self._mkdir(server_deployment_dir)
deployment_path = os.path.join(
server_deployment_dir, d['deployment_name'])
# See if the config can be loaded as a JSON data structure
# In some cases, it may already be JSON (hiera), or it may just
# be a string (script). In those cases, just use the value
# as-is.
try:
data = json.loads(d['config'])
except Exception:
data = d['config']
# If the value is not a string already, pretty print it as a
# string so it's rendered in a readable format.
if not (isinstance(data, six.text_type) or
isinstance(data, six.string_types)):
data = json.dumps(data, indent=2)
d['config'] = data
# The hiera Heat hook expects an actual dict for the config
# value, not a scalar. All other hooks expect a scalar.
if d['group'] == 'hiera':
d['scalar'] = False
else:
d['scalar'] = True
if d['group'] == 'os-apply-config':
message = ("group:os-apply-config is deprecated. "
"Deployment %s will not be applied by "
"config-download." % d['deployment_name'])
warnings.warn(message, DeprecationWarning)
with open(deployment_path, 'wb') as f:
template_data = deployment_template.render(
deployment=d,
server_id=server_ids[server])
self.validate_config(template_data, deployment_path)
f.write(template_data.encode('utf-8'))
# Render group_vars
for role in set(server_roles.values()):
group_var_role_path = os.path.join(group_vars_dir, role)
# NOTE(aschultz): we just use yaml.safe_dump for the vars because
# the vars should already bein a hash for for ansible.
# See LP#1801162 for previous issues around using jinja for this
with open(group_var_role_path, 'w') as group_vars_file:
yaml.safe_dump(role_group_vars[role], group_vars_file,
default_flow_style=False)
# Render host_vars
for server in server_names.values():
host_var_server_path = os.path.join(host_vars_dir, server)
host_var_server_template = env.get_template('host_var_server.j2')
role = server_roles[server]
ansible_host_vars = (
yaml.safe_dump(
role_host_vars[role][server],
default_flow_style=False) if role_host_vars else None)
pre_deployments = server_deployment_names.get(
server, {}).get('pre_deployments', [])
post_deployments = server_deployment_names.get(
server, {}).get('post_deployments', [])
with open(host_var_server_path, 'w') as f:
template_data = host_var_server_template.render(
role=role,
pre_deployments=pre_deployments,
post_deployments=post_deployments,
ansible_host_vars=ansible_host_vars)
self.validate_config(template_data, host_var_server_path)
f.write(template_data)
# Render NetworkConfig data
self.render_network_config(stack, config_dir, server_roles)
shutil.copyfile(
os.path.join(templates_path, 'deployments.yaml'),
os.path.join(config_dir, 'deployments.yaml'))
self.log.info("The TripleO configuration has been successfully "
"generated into: %s" % config_dir)
return config_dir
def download_config(self, name, config_dir, config_type=None,
preserve_config_dir=True, commit_message=None):
if commit_message is None:
commit_message = 'Automatic commit of config-download'
# One step does it all
stack = self.fetch_config(name)
self.create_config_dir(config_dir, preserve_config_dir)
self._mkdir(config_dir)
git_repo = self.initialize_git_repo(config_dir)
self.log.info("Generating configuration under the directory: "
"%s" % config_dir)
self.write_config(stack, name, config_dir, config_type)
self.snapshot_config_dir(git_repo, commit_message)
return config_dir