You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@storm.apache.org by "Jungtaek Lim (JIRA)" <ji...@apache.org> on 2015/11/24 00:30:11 UTC
[jira] [Commented] (STORM-1341) Better to have multilang subprocess
timeout, not relying on supervisor worker timeout
[ https://issues.apache.org/jira/browse/STORM-1341?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15023335#comment-15023335 ]
Jungtaek Lim commented on STORM-1341:
-------------------------------------
Its default value would be SUPERVISOR_WORKER_TIMEOUT_SECS for respecting backward compatibility.
> Better to have multilang subprocess timeout, not relying on supervisor worker timeout
> -------------------------------------------------------------------------------------
>
> Key: STORM-1341
> URL: https://issues.apache.org/jira/browse/STORM-1341
> Project: Apache Storm
> Issue Type: Improvement
> Components: storm-multilang
> Reporter: Jungtaek Lim
> Priority: Minor
>
> Currently, ShellBolt determines that subprocess is hang if there're no message from subprocess within SUPERVISOR_WORKER_TIMEOUT_SECS.
> Since ShellBolt treats all kinds of message to heartbeat (STORM-742), we get rid of heartbeat issue that heartbeat tuple cannot processed in time because too many other BoltMsgs waiting.
> But it doesn't cover two scenario,
> - STORM-758: Subprocess doesn't send anything - emit, ack, fail - so ShellBolt has to rely on heartbeat tuple. It is valid for non-ACK mode.
> - Each (or some kind of) tuple takes very long time to process. For now it should be shorter than SUPERVISOR_WORKER_TIMEOUT_SECS.
> It could be better to have multilang subprocess timeout for clarity.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)