Log rotation

Version added: 3.1.0

logrotate controls the growth of log files on PostgreSQL instance Pods. Without rotation, PostgreSQL and pgBackRest logs can grow until they consume the Persistent Volume and affect database stability. logrotate keeps retention and disk usage predictable.

In Percona Operator for PostgreSQL, logrotate runs in a dedicated logrotate sidecar container on every instance Pod when persistent logging is enabled. It reads the log files at the specified path and rotates them according to the set of rules.

By default, logrotate rotates logs as follows:

• PostgreSQL server *.log and *.csv log files under /pgdata/pgNN/log/:

  • Rotates log files when they reach 100 megabytes in size (size 100M). This acts as a safety measure to prevent the logs from using too much disk space. The rotation by size does not replace or interfere with PostgreSQL’s built-in log rotation, which is based on the age or time of the log files using the logging_collector setting. This ensures both rotation methods work together without causing conflicts.
  • Keeps up to 7 rotated files
  • Uses copytruncate, which means logrotate copies the log file contents to a rotated file and empties the original file in place. This allows processes like PostgreSQL and pgBackRest to keep writing to the same log file without needing to close and reopen it, ensuring uninterrupted logging during rotation.

• pgBackRest client logs on instance Pods under /pgdata/pgbackrest/log/*.log:

  • Rotates daily, or earlier when a file reaches 100 megabytes in size (maxsize 100M)
  • Keeps up to 7 rotated files
  • Uses copytruncate
  • Skips missing or empty log files and leaves rotated logs uncompressed
  • Runs on schedule. By default, this is once per day at midnight.

Configure log rotation

You can customize log rotation if you need different retention, size limits, or want to rotate additional files. Configure it in these ways:

• Override the default logrotate configuration via the Custom Resource
• Add additional configuration via a ConfigMap
• Set a custom log rotation schedule

When you change logrotate configuration, the Operator updates the configuration hash and makes a rolling restart of the instance Pods.

Override the default logrotate configuration

Use the spec.logcollector.logRotate.configuration section in the Custom Resource to completely override the default logrotate settings for the Operator-managed postgres.conf snippet.

Important

You must provide the full logrotate configuration because the Operator replaces the default configuration with the one you provide. Refer to the default configuration to see the built-in rules and use it as a guide for your custom settings.

Here’s an example configuration that rotates PostgreSQL logs hourly and keeps three copies:

spec:
  logcollector:
    enabled: true
    image: docker.io/percona/fluentbit:{{ logcollector }}
    logRotate:
      schedule: "0 * * * *"
      configuration: |
        /pgdata/pg*/log/*.log
        /pgdata/pg*/log/*.csv {
          hourly
          rotate 3
          missingok
          notifempty
          copytruncate
          sharedscripts
        }

Apply the configuration:

kubectl apply -f deploy/cr.yaml -n <namespace>

Extend the default configuration via ConfigMap

You can pass additional logrotate rules via a ConfigMap. Use spec.logcollector.logRotate.extraConfig.name to load additional .conf files from that ConfigMap. The file name must end with .conf.

Important

The postgres.conf key name is reserved for the Operator-managed main configuration. Do not use it in your custom ConfigMap.

For example, you want to record that a log rotation ran. To do that, add a postrotate script to your logrotate configuration. After logrotate finishes rotating the matched files, it appends a timestamped line to a file on the data volume

1. Create a ConfigMap. For example, custom-logrotate.yaml:

   custom-logrotate.yaml

   apiVersion: v1
   kind: ConfigMap
   metadata:
     name: my-logrotate-extra
     namespace: <namespace>
   data:
     extra.conf: |
       /pgdata/pg18/log/*.log {
           daily
           rotate 7
           missingok
           notifempty
           copytruncate
           sharedscripts
           postrotate
               echo "$(date -u +%FT%TZ) rotated on $POD_NAME" >> /pgdata/logrotate-postrotate.log
           endscript
       }      
   

2. Create the ConfigMap:

   kubectl apply -f custom-logrotate.yaml -n <namespace>
   

3. Reference it in the Custom Resource:

   spec:
     logcollector:
       enabled: true
       image: docker.io/percona/fluentbit:{{ logcollector }}
       logRotate:
         extraConfig:
           name: my-logrotate-config
   

4. Apply the Custom Resource:

   kubectl apply -f deploy/cr.yaml -n <namespace>
   

Set a custom schedule

Use spec.logcollector.logRotate.schedule to set the rotation schedule as a cron expression. The default is once per day at midnight (0 0 * * *).

This example runs log rotation every 6 hours:

spec:
  logcollector:
    enabled: true
    image: docker.io/percona/fluentbit:{{ logcollector }}
    logRotate:
      schedule: "0 */6 * * *"

Apply the Custom Resource:

kubectl apply -f deploy/cr.yaml -n <namespace>

---

Last update: September 9, 2026
Created: September 9, 2026