You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@samza.apache.org by "Yi Pan (Data Infrastructure) (JIRA)" <ji...@apache.org> on 2015/09/26 00:40:04 UTC
[jira] [Updated] (SAMZA-778) KafkaCheckpointMigration should not
start/stop the same instance of CoordinatorStreamSystemConsumer multiple
times
[ https://issues.apache.org/jira/browse/SAMZA-778?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yi Pan (Data Infrastructure) updated SAMZA-778:
-----------------------------------------------
Attachment: SAMZA-778-0.patch
Verified by running local migration of 0.9.1 test jobs to 0.10.0 test jobs.
> KafkaCheckpointMigration should not start/stop the same instance of CoordinatorStreamSystemConsumer multiple times
> ------------------------------------------------------------------------------------------------------------------
>
> Key: SAMZA-778
> URL: https://issues.apache.org/jira/browse/SAMZA-778
> Project: Samza
> Issue Type: Bug
> Reporter: Yi Pan (Data Infrastructure)
> Assignee: Yi Pan (Data Infrastructure)
> Attachments: SAMZA-778-0.patch
>
>
> The current implementation of KafkaSystemConsumer only allows start()/stop() once in the life cycle of the same SystemConsumer instance. Unfortunately, KafkaCheckpointMigration code start()/stop() the same instance of KafkaSystemConsumer multiple times, which results in an infinite loop in CoordinatorStreamSystemConsumer.start() in JobRunner.
> The quick fix is to avoid the sequence of SystemConsumer.start()->SystemConsumer.stop()->SystemConsumer.start() in the KafkaCheckpointMigration code.
> In long-term, we may also want to evaluate the use cases where we support start/stop multiple times in the life cycle of a single instance of SystemConsumer.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)