changed instanceSidecarConfiguration not applied

[smiyc]

due to a memory problem I wanted to set the BarmanObjectstore .spec.instanceSidecarConfiguration.logLevel: trace
after applying the logLevel, it was set at the ObjectStore

but the sidecar did not change

after restarting the container, the logLevel was correctly set to trace

[dosubot]

This is a known limitation. The instanceSidecarConfiguration (including logLevel) is read from the ObjectStore spec only at pod creation time via the CNPG-I lifecycle hook and injected as container arguments — it is not watched or dynamically reloaded at runtime [1]. The sidecar starts with --log-level=<value> baked into its args, and once running, there's no mechanism to pick up changes.

Additionally, the ObjectStoreReconciler that could potentially trigger pod restarts on spec changes is currently unimplemented — it's an empty placeholder with TODO comments [2]. This means changes to instanceSidecarConfiguration won't reliably trigger a pod restart either.

This matches issue #817, which reports the same behavior for resource limit changes [3].

Workaround: After changing the ObjectStore spec, trigger a rolling restart with kubectl cnpg restart <cluster-name> to force pods to be recreated with the updated configuration.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}