Open SatyaKuppam opened 1 year ago
@pbadani looping you on this, since you are the last committed user.
Same for me, had to revert to previous Assignor.
Furthermore CooperativeStickyAssignor
will became the defaut assignor one day (KIP about it https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=177048248)
Same for me, had to revert to previous Assignor. Furthermore
CooperativeStickyAssignor
will became the defaut assignor one day (KIP about it https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=177048248)
Hey @BDeus, thanks for confirming. We are using the patch provided in this PR and it working fine with CooperativeStickyAssignor
, we haven't seen issues as yet.
This is also affecting us, and it would be great to have this fixed.
Problem
To decrease the impact of rebalances during rolling bounces of k8s pods, we changed the
partition.assignment.strategy
from the defaultRangeAssignor
toCooperativeStickyAssignor
. After this change we encountered NPEs and theS3SinkTask
goes into an unrecoverable state. We did not find the same issue withStickyAssignor
however.Example of an NPE (this is with
v10.0.7
):Possible Fix in #648