[ https://issues.apache.org/jira/browse/BEAM-3696?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16411156#comment-16411156
]
Jean-Baptiste Onofré commented on BEAM-3696:
--------------------------------------------
Agree, I started to change that: it's also the same for {{JmsIO}} where I started to improve
the behavior.
> MQTT IO should compute watermark and ack messages outside of finalizeCheckpoint method
> --------------------------------------------------------------------------------------
>
> Key: BEAM-3696
> URL: https://issues.apache.org/jira/browse/BEAM-3696
> Project: Beam
> Issue Type: Bug
> Components: io-java-mqtt
> Affects Versions: 2.2.0
> Environment: - Flink - beam-runners-flink_2.10:2.2.0
> - Beam and related jars - 2.2.0
> Reporter: Maxim Kolchin
> Assignee: Jean-Baptiste Onofré
> Priority: Major
>
> I'm experiencing a situation when an incoming message isn't acknowledged (therefore in
sometime broker resend it) and the watermark is not updated while new messages are coming
continuously.
> After some time I've discovered that this situation is related to the fact that finalizaCheckpoint
is not being called.
> I took a look at the Pubsub IO implementation and found that they expect such situation
and do not compute watermark and ack messages in finalizeCheckpoint. Here is the comment about
that: [https://github.com/apache/beam/blob/master/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/pubsub/PubsubUnboundedSource.java#L289]
> Should MQTT IO do the same?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
|