Housekeeping
Housekeeping keeps your nodes tidy and consistent. It does two things: it reconciles drift between what Miabi expects and what Docker is actually running, and it reclaims disk by pruning unused images, containers, and volumes.
Drift reconciliation
Miabi maintains a desired state for every workload — which containers should exist, on which node, with which configuration. Over time the real Docker state can drift from that desired state: a container was stopped by hand, a node restarted, or something was changed outside Miabi.
Reconciliation compares desired state against the live Docker state on each node and reports the differences, such as:
- Containers Miabi expects but that are missing or stopped.
- Containers running with a configuration that no longer matches.
- Resources that exist in Docker but are unknown to Miabi.
You can review the detected drift and let Miabi reconcile — recreating or restarting what's missing so the node matches the desired state again.
Resources that exist in Docker but aren't managed by Miabi are surfaced here too. To bring them under management instead of treating them as drift, use Docker import.
Reclaiming disk
Long-running Docker hosts accumulate dangling images, stopped containers, and orphaned volumes that consume disk. Housekeeping lets you prune these to reclaim space:
| Prune target | What it removes |
|---|---|
| Images | Dangling and unused images not referenced by any managed workload |
| Containers | Stopped containers no longer tracked by Miabi |
| Volumes | Orphaned volumes not attached to any managed resource |
Volume pruning permanently deletes data in unattached volumes. Review the list before pruning, and make sure anything you still need is either attached to a managed workload or backed up.
Running housekeeping
Housekeeping is a platform-admin task (see Platform Administration). You can:
- Run it on demand per node from the Housekeeping panel.
- Review before applying — Miabi shows what it found before you confirm reconciliation or pruning.
Run a reconcile after any manual Docker work or node reboot, and prune images periodically on build-heavy nodes to keep disk usage in check.