Aswin Shakil created HDDS-13614:
-----------------------------------
Summary: Create separate bounded executor for
ClosePipelineCommandHandler and CreatePipelineCommandHandler
Key: HDDS-13614
URL: https://issues.apache.org/jira/browse/HDDS-13614
Project: Apache Ozone
Issue Type: Bug
Reporter: Aswin Shakil
Right now, we have a single unbounded executor for both
ClosePipelineCommandHandler and CreatePipelineCommandHandler. If either of
these handlers is slow, it will cause a backlog on the entire thread's queue.
It will later result in OutOfMemoryError as the queue size keeps on increasing
and crash the DN.
We have a "hdds.datanode.command.queue.limit", but we don't use it for this
executor. We should make this queue bounded and we should also have separate
executors for both so that one handler doesn't become a bottleneck for the
other.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]