Thanks for the KIP, Colin. The KIP cals one metric "NewActiveControllersCount" but we don't append "Count" to the other metric names (e.g. it is "TimedOutBrokerHeartbeats" instead of "TimedOutBrokerHeartbeatsCount"). Should we be consistent (either use the suffix everywhere or don't use it anywhere)?
The phrase "Note that only active controllers handle heartbeats, so only they will see increases in this metric." could apply elsewhere but is not mentioned elsewhere. For example, while standy controllers have a queue and will publish metrics related to it, the queue will remain empty (and the number of timeouts will remain constant) until a controller becomes active. Or is that incorrect? I wonder if we can be more generic and index these things by operation type. Ron On Thu, Jun 1, 2023 at 6:13 PM Colin McCabe <cmcc...@apache.org> wrote: > > Hi all, > > I posted a KIP to add some more metrics for measuring KRaft performance. Take > a look at: https://cwiki.apache.org/confluence/x/gBU0Dw > > best, > Colin
