[
https://issues.apache.org/jira/browse/KAFKA-4793?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15884375#comment-15884375
]
Ewen Cheslack-Postava commented on KAFKA-4793:
----------------------------------------------
To clarify for anyone that happens upon this, it seems like there is some
confusion about the functionality the endpoint provides. In Connect there are
effectively two concepts associated with connectors: the connector instance
which is assigned to a single worker and monitors the input/output system to
determine if any task reconfiguration needs to happen and the logical
connector, which includes that monitoring thread as well as all tasks
associated with the connector (and therefore is distributed across many
workers).
The current functionality of this endpoint is to restart *just* the thread that
monitors the input/output system. It sounds like the request here is that we
should restart all associated tasks as well (giving an easy way to bounce
everything related to a connector rather than having to list out all the tasks
and then bounce them one by one).
The best short term solution we could add and be backwards compatible would be
to add a query parameter/request body flag to use this alternative behavior. We
could then change the default in a future major release version.
> Kafka Connect: POST /connectors/(string: name)/restart doesn't start failed
> tasks
> ---------------------------------------------------------------------------------
>
> Key: KAFKA-4793
> URL: https://issues.apache.org/jira/browse/KAFKA-4793
> Project: Kafka
> Issue Type: Improvement
> Reporter: Gwen Shapira
>
> Sometimes tasks stop due to repeated failures. Users will want to restart the
> connector and have it retry after fixing an issue.
> We expected "POST /connectors/(string: name)/restart" to cause retry of
> failed tasks, but this doesn't appear to be the case.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
