You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@cassandra.apache.org by onmstester onmstester <on...@zoho.com> on 2022/05/09 10:51:44 UTC

increased crossNodeDrop after upgrading from 3.11.2 to 3.11.12

Hi, 



Just upgraded Cassandra cluster from 3.11.2 to 3.11.12, almost everything's fine, except:



In a few minutes interval, one or two nodes reporting a few crossNodeDrop and the node changes every time

Read latency increased by 50%




I noticed multiple gc ParNew logs every minutes on all nodes since the upgrade (tested with heap 16g/g1 and 8g/cms).


Some conclusions about upgrading to 3.11.12:

Average CPU usage decreased by 10% (now is about 50%). 

NTP block is 0 all the time (on previous version it was > 0 always).

In 3.11.2 i've got 18K NTP and 30K Mutation so considering rf=3 the equation of NTP * RF = Mutation, across cluster was not applied, but with the new version it is (so it seems like that all writes fully replicated now?).



How do i find out the root cause of crossNodeDrops?