Nandakumar created HDDS-9479:
--------------------------------
Summary: Pipeline close doesn't wait for containers to be closed
Key: HDDS-9479
URL: https://issues.apache.org/jira/browse/HDDS-9479
Project: Apache Ozone
Issue Type: Improvement
Components: SCM
Affects Versions: 1.3.0
Reporter: Nandakumar
Assignee: Nandakumar
Whenever we close a pipeline, we have an option to give some grace time for the
container to get closed, the grace time is configured using
_ozone.scm.pipeline.destroy.timeout_.
We wait for the timeout to happen before we go ahead and delete the pipeline.
This will give enough time for the datanodes to close the container gracefully.
It will prevent the containers from moving to the QUASI_CLOSED state.
This functionality is broken and we don't wait for the timeout to happen before
we delete the Pipeline. This creates a lot of QUASI_CLOSED containers in the
cluster when a node goes stale or when a datanode is getting decommissioned.
This has to be fixed and we should wait for the configured amount of time
before we delete the Pipeline, this will give datanodes enough time to CLOSE
the containers on that pipeline.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
