Tool to migrate existing data to db per tenant

At present, all time series are accumulated in the same database in
InfluxDB. Separate database per tenant would make queries faster for
tenants. It would also allow administrators to define retention policy
per tenancy.

This changeset adds a migration tool and a guide to allow users to
migrate existing data to a database per tenant.

Task: 36435
Story: 2006331

Change-Id: I3faf3d96b1121c8b8c383dee2fda27d34af5c853
This commit is contained in:
Bharat Kunwar 2019-09-03 15:09:42 +01:00
parent 774e981f87
commit 0daa8ebc76
3 changed files with 361 additions and 2 deletions

View File

@ -45,7 +45,7 @@ def _get_config_files():
return conf_files
def parse_args():
def parse_args(description='Persists metrics & alarm history in TSDB'):
global _CONF_LOADED
if _CONF_LOADED:
LOG.debug('Configuration has been already loaded')
@ -58,7 +58,7 @@ def parse_args():
project='monasca',
version=version.version_str,
default_config_files=_get_config_files(),
description='Persists metrics & alarm history in TSDB')
description=description)
log.setup(CONF,
product_name='monasca-persister',

View File

@ -0,0 +1,122 @@
migrate-to-db-per-tenant.py
===========================
The general plan for the monasca project is to move in the direction of
having a database per tenant because:
- Not only give a finer grain control over retention policy per tenant
but also possibly speed up tenants queries by scoping them within their
project.
- Security is improved though better isolation. For example, a previous bug in
InfluxDB where the tenant ID was ignored in the query exposed data from
outside a tenants project. This is less likely to happen with a separate DB
per tenant.)
- We move in a direction of improving scalability for InfluxDB users
without the Enterprise license. In the future a dedicated InfluxDB
instance could optionally be used per project.
For further reading - https://storyboard.openstack.org/#!/story/2006331
All effort has been made to ensure this is a safe process. And it
should be safe to run the tool multiple times. However, it is provided
AS IS and you should use it at your own risk.
Usage
=====
Steps to use this tool:
- Log in to one node where monasca-persister is deployed.
- Identify installation path to monasca-persister. This may be a
virtual environment such as
`/opt/stack/venv/monasca-<version>/lib/python2.7/site-packages/monasca_persister`
or as in devstack
`/opt/stack/monasca-persister/monasca_persister/` which you may need to
activate.
- Identify the existing configuration for monasca-persister. The likely
location for this is `/etc/monasca/persister.conf` if using devstack.
- Optionally, make a backup of your database in case something goes
wrong but your original database should be left intact.
- Open and edit `migrate-to-db-per-tenant.py` and edit mapping between
your existing projects to retention policies if you so wish. You may
also want to change `end_time_offset_override` to the length of history
you want the migration tool to consider.
- Invoke the tool to migrate to database per tenant. The arguments inside the
square bracket default to the values shown when undefined.
::
sudo -u mon-persister python migrate-to-db-per-tenant.py --config-file /etc/monasca/persister.conf --migrate-retention-policy project:2w,project2:1d --migrate-skip-regex ^log\\..+ [--migrate-time-unit w --migrate-start-time-offset 0 --migrate-end-time-offset 520]
- The progress of the migration will be logged to persister log file
specifed in the persister config.
FAQ
===
1. Will this interrupt the operation of Monasca?
- In theory, you can run this migration query on a live Monasca system
because the migration leaves the original database untouched.
However, in practice, it may degrade performance while the migration
is taking place.
2. How do I ensure that I migrate *all* metrics to the new scheme? i.e.
Do I need to stop all persisters writing to InfluxDB and let Kafka
buffer for a bit while the migration is taking place?
- If you stop the persister and the migration takes longer than the
Kafka buffer duration, this may cause data loss. It may be best to do
the first batch of migration on the entire dataset and perform
iterative migration of smaller chucks collected in the duration
migration was taking place.
3. Will the original index be left intact so that I can fall back if it
doesn't work?
- Yes.
4. What do I do after I have run the tool? I need to enable the feature
before starting the persisters right?
- If you enable the feature by flipping on `db_per_tenant` flag before
starting the migration and restart your persister service, your
persister will start writing to a database per tenant model. It may
be best to switch this feature on after the first round of migration,
restart the persister so that it starts writing to database per
tenancy and migrate the remainder of the data. Note that in the
second run, you may need to delete `migrate-success` so that the metrics
are not skipped. You may also want to try with smaller `--migrate-time-unit`
option (default: `w` for week) if larger chunks of migration fail. Other
option are `d` for day and `h` for hours.
5. How long does it take (rough estimate per GB of metrics?)
- This depends on your storage backend. On one of the systems in which
this tool was tested on, it was an overnight job to move 40GB of metrics.
6. Is the tool idempotent? i.e. could I migrate the bulk of the data
with the persisters running, stop the persisters, and then migrate
the delta in a short interval to minimise downtime? Can I run it
again if it fails?
- Yes, InfluxDB ensures that same time indices with the same tags,
fields and values are not duplicated. Copying things over twice is
perfectly safe.
7. Will it set fire to my cat?
- Depends if you cat is well behaved.
License
=======
Copyright (c) 2019 StackHPC Limited
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.

View File

@ -0,0 +1,237 @@
# (C) Copyright 2019 StackHPC Limited.
#
# 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.
""" Monasca InfluxDB tool for migration to database per tenant
Used to move data from monolithic database e.g. `monasca` to a database per
tenant model, e.g. `monasca_<tenant_id>`.
Please see the included README.rst for more details about creating an
appropriate configuration file.
"""
import os
import re
import sys
from monasca_persister import config
from monasca_persister.repositories.influxdb import metrics_repository
from oslo_config import cfg
from oslo_log import log
LOG = log.getLogger(__name__)
MIGRATE_QUERY = ('SELECT * INTO "{target_db}"..:MEASUREMENT'
' FROM "{measurement}"'
' WHERE _tenant_id=\'{tenant_id}\''
' AND time > {lower_time_offset}'
' AND time <= {upper_time_offset}'
' GROUP BY *')
class MigrationHelper(object):
def __init__(self):
repo = metrics_repository.MetricInfluxdbRepository()
self.conf = repo.conf
self.client = repo._influxdb_client
self.client.switch_database(self.conf.influxdb.database_name)
def _migrate(self, measurement, tenant_id, start_time_offset,
end_time_offset, retention_policy={}, time_unit='w',
db_per_tenant=True, **kwargs):
total_written = 0
first_upper_time_offset = None
last_lower_time_offset = None
time_offset = start_time_offset
if db_per_tenant:
target_db = "{}_{}".format(self.conf.influxdb.database_name, tenant_id)
self.client.create_database(target_db)
if retention_policy:
self.client.create_retention_policy(database=target_db, **retention_policy)
LOG.info(' into {}:'.format(target_db))
while end_time_offset > 0 and time_offset < end_time_offset:
lower_time_offset = 'now()-{}{}'.format(time_offset + 1, time_unit)
upper_time_offset = 'now()-{}{}'.format(time_offset, time_unit)
if not first_upper_time_offset:
first_upper_time_offset = upper_time_offset
migrate_query = MIGRATE_QUERY.format(
target_db=target_db,
measurement=measurement,
tenant_id=tenant_id,
lower_time_offset=lower_time_offset,
upper_time_offset=upper_time_offset,
)
LOG.debug(migrate_query)
written = next(self.client.query(migrate_query).get_points('result')).get('written')
total_written += written
time_offset += 1
if written > 0:
last_lower_time_offset = lower_time_offset
LOG.info(" migrated {} entries from {} -> {} (cumulative {})".format(
written,
lower_time_offset,
upper_time_offset,
total_written,
))
LOG.info(" finished migrating a total of {} entries from {} -> {}.".format(
total_written,
last_lower_time_offset,
first_upper_time_offset,
))
def get_measurements(self, fname):
measurements = []
if fname:
with open(fname, 'a+') as f:
measurements = [l.strip() for l in f.readlines()]
if not measurements:
result = self.client.query('SHOW MEASUREMENTS').get_points('measurements')
measurements = [m.get('name') for m in result]
if fname:
with open(fname, 'w') as f:
for r in measurements:
f.write(r + '\n')
return measurements
def get_tenancies(self, measurements):
result = self.client.query("SHOW TAG VALUES WITH KEY = _tenant_id")
return {m: [t.get('value') for t in result.get_points(m)] for m in measurements}
def get_complete(self, fname):
if fname:
with open(fname, 'a+') as fd:
return {l.strip() for l in fd.readlines()}
else:
return {}
def migrate(self,
tenant_defaults={},
default_start_time_offset=0, # Default: now
default_end_time_offset=(520), # Default: 10 years
skip_regex=[],
measurements_file=None, success_file=None, failure_file=None, **kwargs):
measurements = self.get_measurements(measurements_file)
tenancy = self.get_tenancies(measurements)
done = self.get_complete(success_file)
skip = set()
fail = set()
if failure_file:
if os.path.exists(failure_file):
os.remove(failure_file)
filtered_measurements = []
for measurement in measurements:
if any([f.match(measurement) for f in skip_regex]):
skip.add(measurement)
LOG.debug('Skipping {} because it matches a skip regex.'.format(measurement))
continue
elif measurement in done:
LOG.debug('Skipping {} because its already done.'.format(measurement))
continue
else:
filtered_measurements.append(measurement)
for i, measurement in enumerate(filtered_measurements):
LOG.info('Migrating {}'.format(measurement))
try:
for tenant_id in tenancy.get(measurement):
start_time_offset = tenant_defaults.get(
tenant_id, {}).get('start_time_offset_override',
default_start_time_offset)
end_time_offset = tenant_defaults.get(
tenant_id, {}).get('end_time_offset_override',
default_end_time_offset)
retention_policy = tenant_defaults.get(
tenant_id, {}) .get('rp', {})
self._migrate(measurement, tenant_id,
start_time_offset=start_time_offset,
end_time_offset=end_time_offset,
retention_policy=retention_policy, **kwargs)
if success_file:
with open(success_file, 'a+') as fd:
fd.write('{}\n'.format(measurement))
done.add(measurement)
except Exception as e:
LOG.error(e)
if failure_file:
with open(failure_file, 'a+') as fe:
fe.write('{}\t{}\n'.format(measurement, e))
fail.add(measurement)
LOG.info("{}/{} (done {} + skip {} + fail {})/{}".format(
i + 1, len(filtered_measurements), len(done), len(skip),
len(fail), len(measurements)))
def main():
CONF = cfg.CONF
cli_opts = [
cfg.StrOpt('migrate-time-unit', choices=['h', 'd', 'w'], default='w',
help='Unit of time, h=hour, d=day, w=week (default: "w").'),
cfg.IntOpt('migrate-start-time-offset', default=0,
help='Start time offset in the given unit of time (default: 0).'),
cfg.IntOpt('migrate-end-time-offset', default=520,
help='End time offset in the given unit of time (default: 520).'),
cfg.DictOpt('migrate-retention-policy', default={},
help=('Custom retention policy for projects in the provided'
'time unit. (e.g. project-id-x:2,project-id-y:4)')),
cfg.ListOpt('migrate-skip-regex', default=[],
help=('Skip metrics that match this comma separated list of regex patterns.'
'(e.g. ^log\\\\..+,^cpu\\\\..+ to skip metrics beginning with log.)')),
]
CONF.register_cli_opts(cli_opts)
config.parse_args("Monasca InfluxDB database per tenant migration tool")
# Configure custom retention policy for your existing projects. For
# example, rp2w is a retention policy of two weeks which we can assign to
# project example-project-id.
tenant_defaults = dict()
for k, v in CONF.migrate_retention_policy.items():
if v.isdigit():
rp = '{}{}'.format(v, CONF.migrate_time_unit)
tenant_defaults[k] = dict(
end_time_offset_override=int(v),
rp=dict(name=rp, duration=rp, replication='1', default=True),
)
LOG.info('Project {} will be applied retention policy: {}.'.format(k, rp))
else:
raise ValueError('Retention policy for project {} must be an'
'integer of given time unit. Current value:'
'{}.'.format(k, v))
skip_regex = []
for p in CONF.migrate_skip_regex:
skip_regex.append(re.compile(str(p)))
LOG.info('Metrics matching pattern "{}" will be skipped.'.format(p))
helper = MigrationHelper()
helper.migrate(skip_regex=skip_regex,
tenant_defaults=tenant_defaults,
default_end_time_offset=CONF.migrate_end_time_offset,
default_start_time_offset=CONF.migrate_start_time_offset,
time_unit=CONF.migrate_time_unit,
measurements_file='migrate-measurements',
success_file='migrate-success',
failure_file='migrate-failure')
return 0
if __name__ == "__main__":
sys.exit(main())