janhoy commented on issue #724: URL: https://github.com/apache/solr-operator/issues/724#issuecomment-6765981566
🤖 Took another look at this one with Claude, and I think we can get it to a conclusion. **1.** The failed-scale-down path from @HoustonPutman's last comment: checked the code, it isn't reachable. The `UndoFailedScaleDown` balance requires a failed `ScaleDownLock` sitting in the retry queue, and that lock is only ever taken when `vacatePodsOnScaleDown` isn't `false`. Are you sure you did not use an **ephemeral** cloud @sigram ? **2.** Docs: `cluster-operations.adoc` now lists both balance triggers, which covers part of the original ask, but we should spell out in `scaling.adoc`/`managed-updates.adoc` that the scaling options don't apply to rolling updates on ephemeral clouds, and move the async/retry notes out of the `BalanceReplicasForCluster` comment into the docs as @sigram asked. **3.** Add a narrow opt-out: `spec.scaling.balanceReplicasOnUpdate` (default `true`), suppressing only the post-rolling-update balance. Keeping eviction and balancing separate matters here — evicting before deleting an ephemeral pod is data safety, while the final balance is placement optimization and exactly what you'd want to own yourself if you manage replicas externally. Small change: one CRD field, one condition in `handleManagedCloudRollingUpdate`, docs, and an e2e spec asserting no `BalanceReplicas` op is queued. @sigram would 3 cover your case, or do you also need the per-pod eviction during rolling updates turned off? That one's a bigger call since skipping it drops the replicas on that pod... -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
