[
https://issues.apache.org/jira/browse/STORM-742?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14482674#comment-14482674
]
ASF GitHub Bot commented on STORM-742:
--------------------------------------
Github user dashengju commented on the pull request:
https://github.com/apache/storm/pull/497#issuecomment-90392638
@HeartSaVioR
We have tested the patch. It solved most of the problem, but it still exist
in some cases.
Currently, even with ACK, the subprocess handle one tuple, emit the
result, and read all the tuples from stdin to get the id, read all the tuples
from stdin needs long time, which cause exceed timeout.
So we limit the queue-size in ShellBolt to 100, just make the read all the
tuples from stdin operation a short time can avoid the problem.
> Very busy ShellBolt subprocess with ACK mode cannot respond heartbeat just in
> time
> ----------------------------------------------------------------------------------
>
> Key: STORM-742
> URL: https://issues.apache.org/jira/browse/STORM-742
> Project: Apache Storm
> Issue Type: Bug
> Affects Versions: 0.9.3, 0.10.0, 0.9.4, 0.11.0
> Reporter: Jungtaek Lim
> Assignee: Jungtaek Lim
> Priority: Critical
>
> As [~dashengju] stated from STORM-738, very busy ShellBolt subprocess cannot
> respond heartbeat just in time.
> Actually it's by design constraint (more details are on STORM-513 or
> STORM-738), but ShellSpout avoids constraint by updating heartbeat at any
> type of response from subprocess.
> We can apply this approach to ShellBolt and let ShellBolt avoid design
> constraint, too.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
