You are viewing a plain text version of this content. The canonical link for it is here.

Posted to commits@beam.apache.org by "Axel Magnuson (JIRA)" <ji...@apache.org> on 2018/03/30 21:20:00 UTC

[jira] [Commented] (BEAM-3976) SdkHarnessClient is thread-safe

    [ https://issues.apache.org/jira/browse/BEAM-3976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16420943#comment-16420943 ] 

Axel Magnuson commented on BEAM-3976:
-------------------------------------

After investigating and discussing with Thomas Groh, this has already been fixed. the SdkHarnessClient should be threadsafe.  See line 57 in:

 

[https://github.com/apache/beam/blob/master/runners/java-fn-execution/src/main/java/org/apache/beam/runners/fnexecution/control/FnApiControlClient.java]

 

 

> SdkHarnessClient is thread-safe
> -------------------------------
>
>                 Key: BEAM-3976
>                 URL: https://issues.apache.org/jira/browse/BEAM-3976
>             Project: Beam
>          Issue Type: Bug
>          Components: runner-core
>            Reporter: Ben Sidhom
>            Assignee: Axel Magnuson
>            Priority: Major
>
> In general, we want to share access to a given SDK harness among multiple runner workers as a way to amortize container startup and resource costs. Because control messages are multiplexed over the same shared control connection, this sharing currently requires external locking on the same shared lock object. This is error-prone and difficult to verify. The SdkHarnesClient should use internal mechanisms to provide thread-safety.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)