Details
-
Bug
-
Resolution: Won't Do
-
Major
-
5.0.0
-
Untriaged
-
-
Unknown
Description
Build : 5.0.0-3509
On the system testing cluster that runs longevity with Ephemeral buckets, the memory consumption for a few buckets is unusually high.
DISTRICT bucket has 50 items, but the memory used is 757MB.
default bucket has 2000 items, but the memory used is 11.6GB.
We had created a primary index on DISTRICT bucket to check the data size, and the index data size was ~6KB. This suggests that the compactor might not be working properly and the memory is not freed up.
The cluster is live at http://172.23.106.14:8091 (logs attached)
A similar issue is also seen in another non-ephemeral buckets longevity on the same build. It is observed that the resident ratios for some buckets are unusually high.
DISTRICT bucket has 82605 items and the bucket has 736MB RAM allocated, but the resident ratio is 0%
default bucket has 931000 items, bucket memory is 18GB, resident ratio is 16.1% (item size is ~1K max)
This cluster is live at http://172.23.108.103:8091