Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.5.0
-
Untriaged
-
Centos 64-bit
-
-
No
Description
Build 5.5.0-2414.
Mapping the number of writer threads to the number of CPU cores is not always a good idea. I attempted to create ten partitioned indexes on my 48-core server and the indexer process was OOM-killed almost immediately.
Although it's possible to mitigate this issue by tuning indexer.numSliceWriters option, a global thread pool with fair scheduling seems more appropriate.
MB-29009 describes the environment and secondary indexes.