Cleanup
Images pile up with every deployment, build caches grow, and HivePaaS's records of old tasks and deployments add up. A cleanup job clears them on a schedule.
The cleanup job
Settings → Data Cleanup sets it:
| Options | What it clears |
|---|---|
| DB Cleanup Options | HivePaaS's records, once older than their retention: tasks, deployments, audit logs, system errors, deleted objects. |
| File Cleanup Options | Files HivePaaS no longer needs. |
| Cache Cleanup Options | Build caches and repository caches, past their retention. |
| Docker Swarm Cleanup Options | On every node: stopped containers, unused images, networks and volumes, and the build cache, each if chosen. |
It runs on its schedule: by default every day at midnight in the installation's timezone, the one the installer asked for. Its schedule is a Cron Expression or an interval. Run Cleanup Now runs it at once. A notification can report each run.
By default, it keeps tasks, deployments, audit logs, system errors and deleted objects for 90 days, repository caches for 10 days, and the build cache for 30 days.
It also removes apps left behind by a project or environment that was deleted.
Prune Volumes deletes anonymous volumes no container uses. Named volumes, such as apps' storage, are kept.
System apps
With System Apps → Sync With Settings, on by default, each run also brings the apps HivePaaS runs for itself - the registry, and the logging backend and collector - to their settings in System → Registry and System → Logging:
- One switched off is removed. Its data - the registry's images, the stored logs - is kept.
- One switched on but missing is created, and one whose service is gone is created again, with its data.
- One running an image or settings other than its own is deployed again.
What a deployment would not mend is reported, not changed: an app scaled to zero, or one whose containers keep failing. The run then fails, so that its notification says what needs looking at; its log has the details.
OBI is checked too, on the nodes System → Logging lists for performance data. A node no longer in the cluster is taken off the list. A node whose agent is not running, or where OBI cannot run - an old kernel, too little memory - is reported. Nothing is synced while HivePaaS is being updated.
Clearing caches now
Force Clear Build Cache and Force Clear Repo Cache, on the same page, clear those at once, such as when a build keeps using an outdated layer.
Backup repositories
Backup repositories are cleaned apart, by the Backup Repo Cleanup job: see Backup repositories.