Follow-up from #16237 (per @AnatolyPopov review).
On overloaded Connect clusters, partial commit failures (triggered by commitState.isCommitTimedOut()) can become dominant. Operators should have visibility into how often this happens.
Proposal: expose a counter metric for partial commit failures that operators can alert on.
Willingness to contribute
Follow-up from #16237 (per @AnatolyPopov review).
On overloaded Connect clusters, partial commit failures (triggered by
commitState.isCommitTimedOut()) can become dominant. Operators should have visibility into how often this happens.Proposal: expose a counter metric for partial commit failures that operators can alert on.
Willingness to contribute