Stale Catalog Entries Persist After Refresh: Is Automatic Cleanup Supported?

Question

Stale Catalog Entries After Refresh
When a Rancher catalog source is refreshed via the catalog.refresh API, entries that have been deleted in the upstream repository are not automatically removed from the Rancher catalog. The catalog list continues to display these stale items until a subsequent manual or automatic cleanup step occurs.
Documentation for Rancher 2.x does not specify an automatic cleanup policy or a configurable flag to enforce removal of such entries. Community discussions note that the absence of automatic cleanup can lead to confusion and clutter in the catalog UI.
Key uncertainties remain about the behavior across Rancher releases and whether any hidden setting or future change will alter this process.
Questions:

Does Rancher automatically delete catalog entries that are no longer present in the source after a refresh?
Is there a configuration option or flag that can enable or disable this cleanup behavior?
Has the behavior changed in newer Rancher releases (e.g., 2.10+)?

Tasadduq Burney · Answer

Answer to the core question
Rancher does **not** automatically delete catalog entries that have been removed from the upstream repository when you run a refresh. The refresh operation only adds or updates charts; it never compares the existing catalog with the source to prune missing items.
When can stale entries be removed?

UI‑initiated refresh – Starting with Rancher 2.10 the refresh dialog includes a checkbox labeled “Prune stale entries”. If you check this box before clicking Refresh, the controller will delete any entries that are no longer present in the source.
  API‑initiated refresh – The /v3/catalogtemplates?action=refresh endpoint accepts a JSON payload. Setting { "prune": true } triggers the same prune behavior; omitting the field or setting it to false leaves stale entries in place.
  No global toggle – There is no Helm chart value, environment variable, or catalog‑source spec that changes the default behavior to always prune. The default remains “no‑prune” unless the prune flag is explicitly supplied per‑refresh.

Verification steps for your environment

Confirm the Rancher version (e.g., via UI footer or kubectl get rancher -o jsonpath='{.status.version}'). The behavior described applies to Rancher 2.x up to the latest 2.12‑series; versions prior to 2.10 lack the UI prune checkbox.
  Identify a chart that has been deleted from the upstream Helm repo or Git repository.
  Perform a refresh:
    
      Via UI: leave the “Prune stale entries” box unchecked, then note whether the deleted chart still appears.
      Via API: call curl -k -u "$TOKEN" -X POST "https:///v3/catalogtemplates?action=refresh" -H 'Content-Type: application/json' -d '{}' and check the catalog list.
    
  Repeat the refresh with the prune option enabled (UI checkbox checked or API payload { "prune": true }) and verify that the chart disappears.

What to ask next
If you are using a Rancher version older than 2.10, the UI does not expose a prune option at all; in that case the only reliable way to remove stale entries is to delete and re‑add the catalog source. Knowing your exact Rancher version will confirm whether the UI checkbox or API prune flag is available to you.

Stale Catalog Entries Persist After Refresh: Is Automatic Cleanup Supported?

Stale Catalog Entries After Refresh

1 answer

Answer to the core question

When can stale entries be removed?

Verification steps for your environment

What to ask next

Your perspective belongs here.

0 question comments