[
https://issues.apache.org/jira/browse/CAMEL-22173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17981675#comment-17981675
]
Claus Ibsen commented on CAMEL-22173:
-------------------------------------
As you have the problem now you are welcome to see if you can improve this and
send suggested code patches.
> camel-salesforce - Checking initial replay id takes a lot of time or hangs
> --------------------------------------------------------------------------
>
> Key: CAMEL-22173
> URL: https://issues.apache.org/jira/browse/CAMEL-22173
> Project: Camel
> Issue Type: Bug
> Components: camel-salesforce
> Affects Versions: 4.10.4
> Reporter: Lorenzo Benvenuti
> Priority: Major
>
> Hi,
> we've just updated Camel from 4.4 to 4.10 and we've noticed an issue which is
> likely caused by the changes introduced by CAMEL-21740. We're seeing two
> issues:
> - When a consumer restarts from the latest event (i.e. all the events have
> already been consumed) then checking the replay id takes exactly 60s. I think
> this happens because {{PubSubApiClient}} sets {{FetchRequest.numRequested}}
> to 1, but there aren't any events to consume; in this scenario the client
> waits up to 60s before invoking the callback. I'm not sure where this value
> is configured (client side? Server side?). I had a look at the Camel code
> and couldn't find this setting anywhere. I've also tried to pass
> {{numRequested=0}} and it doesn't work (it returns immediately but it doesn't
> check the replay id at all)
> - Sometimes the "check initial replay id" call just hangs. Bumping the
> timeout doesn't seem to help: we're using 5 minutes right now and we're still
> seeing issues. I'm not sure what's triggering this behavior; AFAICT after we
> restart the application the call is always successful.
> Based on the comments in CAMEL-21740 I understand that the purpose of this
> check is to avoid using LATEST if the replay id doesn't exist; I'm wondering
> if this could be done directly when subscribing to the topic, instead of
> trying to fetch a single event beforehand.
> Thanks,
> lorenzo
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
