pglookout

pglookout is a PostgreSQL® replication monitoring and failover daemon. pglookout monitors PG database nodes and their replication status and acts according to that status, for example calling a predefined failover command to promote a new primary in case the previous one goes missing.

pglookout supports two different node types, ones that are installed on the db nodes themselves, and observer nodes that can be installed anywhere. The purpose of pglookout on the PostgreSQL DB nodes is to monitor the replication status of the cluster and act accordingly, the observers have a more limited remit: they just observe the cluster status to give another viewpoint to the cluster state.

A single observer can observe any number of PostgreSQL replication clusters simultaneously. This makes it possible to share an observer between multiple replication clusters. In general it is recommended that you run with at least one external observer giving an additional viewpoint on the health of the cluster.

Requirements

pglookout can monitor PostgreSQL versions 10 and above.

pglookout has been developed and tested on modern Linux x86-64 systems, but should work on other platforms that provide the required modules. pglookout is implemented in Python and works with CPython versions 3.9 or newer. pglookout depends on the Requests and Psycopg2 Python modules.

Building

To build an installation package for your distribution, go to the root directory of a pglookout Git checkout and then run:

Debian:

make deb

This will produce a .deb package into the parent directory of the Git checkout.

Fedora:

make rpm

This will produce a .rpm package into rpm/RPMS/noarch/.

Python/Other:

python setup.py bdist_egg

This will produce an egg file into a dist directory within the same folder.

Installation

To install it run as root:

Debian:

dpkg -i ../pglookout*.deb

Fedora:

dnf install rpm/RPMS/noarch/*

On Linux systems it is recommended to simply run pglookout under systemd:

systemctl enable pglookout.service

and eventually after the setup section, you can just run:

systemctl start pglookout.service

Python/Other:

easy_install dist/pglookout-1.4.0-py3.8.egg

On systems without systemd it is recommended that you run pglookout under Supervisor or other similar process control system.

Setup

After this you need to create a suitable JSON configuration file for your installation.

1. Create a suitable PostgreSQL user account for pglookout:

   CREATE USER pglookout PASSWORD 'putyourpasswordhere';
   

2. Edit the local pg_hba.conf to allow access for the newly created account to the postgres (or other suitable database of your choice) from the primary, replica and possible observer nodes. While pglookout will only need to run a few builtin functions within the database, it is still recommended to setup a separate empty database for this use. Remember to reload the configuration with either:

   SELECT pg_reload_conf();
   

   or by sending directly a SIGHUP to the PostgreSQL postmaster process.

3. Fill in the created user account and primary/replica/observer addresses into the configuration file pglookout.json to the section remote_conns.

4. Create a failover script and add the path to it into the configuration key failover_command. As an example failover script, a shell script that uses IP aliasing is provided in the examples. It is recommended to provide some way to provide STONITH (Shoot The Other Node In The Head) capability in the script. Other common methods of doing the failover and getting DB traffic diverted to the newly promoted primary are the switching of PgBouncer (or other poolers) traffic, or changes in PL/Proxy configuration.

   You should try to run the failover script you provide with pglookout's user privileges to see that it does indeed work.

5. Now copy the same pglookout.json configuration to the replica and possible observer nodes but you need to edit the configuration on the other nodes so that the own_db configuration variable matches the remote_conns key of the node. For observer nodes, you can leave it as an empty '' value, since they don't have a DB of their own.

Other possible configuration settings are covered in more detail under the Configuration keys section of this README.

6. If all has been set up correctly up to this point, pglookout should now be ready to be started.

Alert files

Alert files are created whenever an error condition that requires human intervention to solve. You're recommended to add checks for the existence of these files to your alerting system.

authentication_error

There has been a problem in the authentication of at least one of the PostgreSQL connections. This usually denotes either a wrong username/password or incorrect pg_hba.conf settings.

multiple_master_warning

This alert file is created when multiple primaries are detected in the same cluster.

replication_delay_warning

This alert file is created when replication delay goes over the set warning limit. (this is warning is an exception to the rule that human intervention is required. It is only meant as an informative heads up alert that a failover may be imminent. In case the replication delay drops below the warning threshold again, the alert will be removed)

failover_has_happened

This alert file is created whenever the failover command has been issued.

General notes

If correctly installed, pglookout comes with two executables, pglookout and pglookout_current_master that both take as their arguments the path to the node's JSON configuration file.

pglookout is the main process that should be run under systemd or supervisord.

pglookout_current_master is a helper that will simply parse the state file and return which node is the current primary.

While pglookout is running it may be useful to read the JSON state file that exists where json_state_file_path points. The JSON state file is human readable and should give an understandable description of the current state of the cluster which is under monitoring.

Configuration keys

autofollow (default false)

Do you want pglookout to try to start following the new primary. Useful in scenarios where you have a primary and two replicas, primary dies and another replica is promoted. This will allow the remaining replica to start following the new primary. Requires pg_data_directory, pg_start_command and pg_stop_command configuration keys to be set.

db_poll_interval (default 5.0)

Interval on how often should the connections defined in remote_conns be polled for information on DB replication state.

remote_conns (default {})

PG database connection strings that the pglookout process should monitor. Keys of the object should be names of the remotes and values must be valid PostgreSQL connection strings or connection info objects.

primary_conninfo_template

Connection string or connection info object template to use when setting a new primary_conninfo value for recovery.conf after a failover has happened. Any provided hostname and database name in the template is ignored and they are replaced with a replication connection to the new primary node.

Required when autofollow is true.

observers (default {})

This object contains key value pairs like {"1.2.3.4": "http://2.3.4.5:15000"}. They are used to determine the location of pglookout observer processes. Observers are processes that don't take any actions, but simply give a third party viewpoint on the state of the cluster. Useful especially during net splits.

poll_observers_on_warning_only (default False)

this allows observers to be polled only when replication lag is over warning_replication_time_lag

http_address (default "")

HTTP webserver address, by default pglookout binds to all interfaces.

http_port (default 15000)

HTTP webserver port.

replication_state_check_interval (default 10.0)

How often should pglookout check the replication state in order to make decisions on should the node be promoted.

failover_sleep_time (default 0.0)

Time to sleep after a failover command has been issued.

maintenance_mode_file (default "/tmp/pglookout_maintenance_mode_file")

If a file exists in this location, this node will not be considered for promotion to primary.

missing_master_from_config_timeout (default 15)

In seconds the amount of time before we do a failover decision if a previously existing primary has been removed from the config file and we have gotten a SIGHUP.

alert_file_dir (default os.getcwd())

Directory in which alert files for replication warning and failover are created.

json_state_file_path (default "/tmp/pglookout_state.json")

Location of a JSON state file which describes the state of the pglookout process.

max_failover_replication_time_lag (default 120.0)

Replication time lag after which failover_command will be executed and a failover_has_happened file will be created.

warning_replication_time_lag (default 30.0)

Replication time lag at which point to execute over_warning_limit_command and to create a warning file.

failover_command (default "")

Shell command to execute in case the node has deemed itself in need of promotion

known_gone_nodes (default [])

Lists nodes that are explicitly known to have left the cluster. If the old primary is removed in a controlled manner it should be added to this list to ensure there's no extra delay when making promotion decision.

never_promote_these_nodes (default [])

Lists the nodes that will never be considered valid for promotion. As in if you have primary p which fails and replicas a and `b, even if b is ahead but is listed in never_promote_these_nodes, a will be promoted.

over_warning_limit_command (default null)

Shell command to be executed once replication lag is warning_replication_time_lag

own_db

The key of the entry in remote_conns that matches this node.

log_level (default "INFO")

Determines log level of pglookout.

pg_data_directory (default "/var/lib/pgsql/data")

PG data directory that needs to be set when autofollow has been turned on. Note that pglookout needs to have the permissions to write there. (specifically to recovery.conf)

pg_start_command (default "")

Command to start a PostgreSQL process on a node which has autofollow set to true. Usually something like "sudo systemctl start postgresql".

pg_stop_command (default "")

Command to stop a PostgreSQL process on a node which has autofollow set to true. Usually something like "sudo systemctl start postgresql".

syslog (default false)

Determines whether syslog logging should be turned on or not.

syslog_address (default "/dev/log")

Determines syslog address to use in logging (requires syslog to be true as well)

syslog_facility (default "local2")

Determines syslog log facility. (requires syslog to be true as well)

statsd (default: disabled)

Enables metrics sending to a statsd daemon that supports the StatsD / Telegraf syntax with tags.

The value is a JSON object:

{
    "host": "<statsd address>",
    "port": "<statsd port>",
    "tags": {
        "<tag>": "<value>"
    }
}

The tags setting can be used to enter optional tag values for the metrics.

Metrics sending follows the Telegraf spec.

cluster_monitor_health_timeout_seconds (default: 2 * replication_state_check_interval)

If set, it will increase the statsd counter cluster_monitor_health_timeout if the cluster_monitor thread has not successfully completed a check since cluster_monitor_health_timeout_seconds.

failover_on_disconnect (default true)

Determines if we take a fail-over decision if we're not connected to the primary anymore.

License

pglookout is licensed under the Apache License, Version 2.0. Full license text is available in the LICENSE file and at http://www.apache.org/licenses/LICENSE-2.0.txt

Credits

pglookout was created by Hannu Valtonen & the Ohmu team for F-Secure and is now maintained by Aiven developers <opensource@aiven.io>.

Recent contributors are listed on the GitHub project page, https://github.com/aiven/pglookout/graphs/contributors

Trademarks

Postgres, PostgreSQL and the Slonik Logo are trademarks or registered trademarks of the PostgreSQL Community Association of Canada, and used with their permission.

Debian, Fedora, Python, Telegraf are trademarks and property of their respective owners. All product and service names used in this website are for identification purposes only and do not imply endorsement.

Contact

Bug reports and patches are very welcome, please post them as GitHub issues and pull requests at https://github.com/aiven/pglookout . Any possible vulnerabilities or other serious issues should be reported directly to the maintainers <opensource@aiven.io>.