Stephen O'Donnell created HDDS-11444:
----------------------------------------
Summary: Make Datanode Command metrics consistent across all
commands
Key: HDDS-11444
URL: https://issues.apache.org/jira/browse/HDDS-11444
Project: Apache Ozone
Issue Type: Improvement
Components: Ozone Datanode
Reporter: Stephen O'Donnell
>From PR:
https://github.com/apache/ozone/pull/7140#pullrequestreview-2287063851
{quote}
We still have to work out the differences in how metrics are tracked in between
datanode commands that go through the replication supervisor and those that are
handled outside of it. For example in ReplicateContainerCommandHandler:
* totalTime and invocationCount are never assigned. Their related metrics
will always be 0.
* There is no analogous metric for execution times in the supervisor to
replace these.
* Queued count is pulled from the supervisor and propagated to the command
handler metrics. The supervisor does not expose this metric directly, even
though it tracks it.
Basically the way this change is implemented I think it is confusing whether to
look for a metric in the replication supervisor or the command handler. IMO all
datanode commands should properly fill in a core common set of metrics, and the
replication supervisor metrics should extend that with additional data for
replication specific commands.
The above PR was as significant improvement so it was commited and this Jira
raised to track this additional task.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
