Rajan Dhabalia
Rajan Dhabalia
> The change in #7406 is to avoid any contention in write path and to impose a global max. I have gone through with #7406 PR but as I said...
/pulsarbot run-failure-checks
> Finally, the number of messages/s is not a good proxy measure for the amount of memory required in the broker Yes, I will add restriction based on number bytes...
> Is it the same as #6178 does? Yes, but #6178 uses a volatile counter and that creates contention in write path and therefore, @merlimat tried to remove it in...
@merlimat we need a fix for the OOM soon. I don't see any issue with this PR but if we still want to move forward with #7406 then can we...
We are frequently seeing this issue again and again in our env and it impacts multiple brokers at the same time and they crash. Can we review the PR or...
I will try one more time to comment on this PR. We have been facing this issue in production for a long time. Should this PR come from a specific...
We have added this PR #7499 patch in internally and was working fine but after upgrading pulsar version and without this change our broker again started crashing. The main issue...
> by dynamically adjusting per-thread quotas. how do we train broker to adjust per-thread quotas dynamically?
@merlimat > I still think we shouldn't add contended counters. We have validated that previous OOMs are not happening anymore in 2.8. Even then, I still believe we should go...