[
https://issues.apache.org/jira/browse/JAMES-3062?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17036733#comment-17036733
]
Benoit Tellier commented on JAMES-3062:
---------------------------------------
After discussing with @mbaecler we came to the conclusion that:
- We don't need a count information, that `isEmpty` / not empty allows to give
enough information to the admin to know something is wrong
Then this can be esili computed with a partially paged select in Cassandra:
[SELECT (*) with batchSize=10].findAny().isPresent()
We should then expose a `DeadLetter::isEmpty` method.
Once done, a health check can be wrapped around it. Along side with regular
helthcheck execution and log output, we would ensure that the admin is warned.
> DeadLetter: size gauge
> ----------------------
>
> Key: JAMES-3062
> URL: https://issues.apache.org/jira/browse/JAMES-3062
> Project: James Server
> Issue Type: Improvement
> Reporter: Nguyễn Việt Đức
> Priority: Major
>
> *Why* This allows a grafana board on EventBus fatal errors
> The admin would thus know he needs to go and analyse/reprocess/clean dead
> letter.
> Modify dead-letter with a "Publisher failedEventCount()" method - and write
> unit tests for it.
> Implement it with a counter for "Cassandra Event dead letter".
> DOD: Test gauge size as part of deadLetter contract, all implementation
> should publish it. Add a grafana board for this.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]