coilyco-flight-deck/infrastructure

Watch

The openclaw namespace is completely empty and has been for ~70 days #812

New issue

Closed

opened 2026-08-13 07:07:40 +00:00 by coilyco-ops · 1 comment

coilyco-ops commented

2026-08-13 07:07:40 +00:00

Owner

What this is

A leftover. kubectl get all,pvc -n openclaw:

No resources found in openclaw namespace.

Not one pod, Service, Deployment, or PVC. The namespace object exists and contains nothing.

What is left of it

The only surviving trace is a Released PersistentVolume pointing at a PVC that no longer exists:

PV	Claim	Status	Age
`pvc-e56c913b-a8c9-4b01-ba03-c6dd2d7ca579`	`openclaw/openclaw-home-pvc`	Released	70d

That PV is one of the three stuck in a delete loop in #810 — 1473 failed deletion attempts, with no backing data left on disk. So openclaw is contributing about a third of that event noise for a workload that no longer exists in any form.

The earlier volume scan reported the namespace as 1 volume, 0 bytes, 0 pods, which is consistent: the PVC was deleted, the PV was released, the data was cleaned, and the namespace was never removed.

Why bother

Almost nothing, honestly — an empty namespace costs no resources. Two reasons it is worth closing out:

It is the clearest case in #810. The two forgejo/data-forgejo-runner-* PVs there need care because PVCs of those exact names are live and bound today. This one has no such ambiguity: nothing named openclaw exists anywhere in the cluster.
An empty namespace is indistinguishable from a namespace whose workload failed to schedule. Removing it means the next person reading a namespace list does not have to work out which it is.

openclaw does not appear in any manifest in infrastructure or deploy, consistent with a decommission that removed the workload and stopped there. With prune: false on the Flux path during the soak, nothing was ever going to clean it up on its own.

Acceptance

The openclaw namespace is deleted, or a reason to keep it is recorded.
Its Released PV goes with it, removing one of the three delete loops in #810.

Notes

No changes made. Low severity — filed at Kai's request after I flagged it as a passing observation inside #810 rather than tracking it.

Next owner

Ops.

## What this is A leftover. `kubectl get all,pvc -n openclaw`: ``` No resources found in openclaw namespace. ``` Not one pod, Service, Deployment, or PVC. The namespace object exists and contains nothing. ## What is left of it The only surviving trace is a `Released` PersistentVolume pointing at a PVC that no longer exists: | PV | Claim | Status | Age | | --- | --- | --- | --- | | `pvc-e56c913b-a8c9-4b01-ba03-c6dd2d7ca579` | `openclaw/openclaw-home-pvc` | Released | 70d | That PV is one of the three stuck in a delete loop in https://forgejo.coilysiren.me/coilyco-flight-deck/infrastructure/issues/810 — 1473 failed deletion attempts, with no backing data left on disk. So `openclaw` is contributing about a third of that event noise for a workload that no longer exists in any form. The earlier volume scan reported the namespace as 1 volume, 0 bytes, 0 pods, which is consistent: the PVC was deleted, the PV was released, the data was cleaned, and the namespace was never removed. ## Why bother Almost nothing, honestly — an empty namespace costs no resources. Two reasons it is worth closing out: 1. It is the clearest case in https://forgejo.coilysiren.me/coilyco-flight-deck/infrastructure/issues/810. The two `forgejo/data-forgejo-runner-*` PVs there need care because PVCs of those exact names are live and bound today. This one has no such ambiguity: nothing named `openclaw` exists anywhere in the cluster. 2. An empty namespace is indistinguishable from a namespace whose workload failed to schedule. Removing it means the next person reading a namespace list does not have to work out which it is. `openclaw` does not appear in any manifest in `infrastructure` or `deploy`, consistent with a decommission that removed the workload and stopped there. With `prune: false` on the Flux path during the soak, nothing was ever going to clean it up on its own. ## Acceptance - The `openclaw` namespace is deleted, or a reason to keep it is recorded. - Its `Released` PV goes with it, removing one of the three delete loops in https://forgejo.coilysiren.me/coilyco-flight-deck/infrastructure/issues/810. ## Notes No changes made. Low severity — filed at Kai's request after I flagged it as a passing observation inside https://forgejo.coilysiren.me/coilyco-flight-deck/infrastructure/issues/810 rather than tracking it. ## Next owner Ops.

coilyco-ops commented

2026-08-25 06:10:23 +00:00

Author

Owner

Resolved. The namespace no longer exists at all, rather than existing empty:

$ kubectl get ns openclaw
Error from server (NotFound): namespaces "openclaw" not found

Its only surviving trace was pvc-e56c913b…, the openclaw/openclaw-home-pvc PV stuck Released for 82 days. That was deleted on 2026-08-25 alongside the two retired runner PVs (#588, #810, #863, #901 all covered the same three). Its data directory was already gone from disk, confirmed through the privileged node-stats volume scan.

Nothing named openclaw remains in the cluster. Closing.

Resolved. The namespace no longer exists at all, rather than existing empty: ``` $ kubectl get ns openclaw Error from server (NotFound): namespaces "openclaw" not found ``` Its only surviving trace was `pvc-e56c913b…`, the `openclaw/openclaw-home-pvc` PV stuck `Released` for 82 days. That was deleted on 2026-08-25 alongside the two retired runner PVs (#588, #810, #863, #901 all covered the same three). Its data directory was already gone from disk, confirmed through the privileged node-stats volume scan. Nothing named `openclaw` remains in the cluster. Closing.

coilyco-ops closed this issue

2026-08-25 06:10:24 +00:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

coilyco-flight-deck/infrastructure#812

No description provided.

Rows
Columns