[jira] [Commented] (STORM-1928) ShellSpout should send heartbeat tuple while nextTuple() is not called

["Jungtaek Lim (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/STORM-1928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15346550#comment-15346550 ] 

Jungtaek Lim commented on STORM-1928:
-------------------------------------

ShellSpout should be changed to alike ShellBolt, I mean having another writer thread, but spout methods should respect blocking manner (which is how spout runs) which makes problem a bit complicated.

> ShellSpout should send heartbeat tuple while nextTuple() is not called
> ----------------------------------------------------------------------
>
>                 Key: STORM-1928
>                 URL: https://issues.apache.org/jira/browse/STORM-1928
>             Project: Apache Storm
>          Issue Type: Bug
>          Components: storm-multilang
>            Reporter: Jungtaek Lim
>            Assignee: Jungtaek Lim
>            Priority: Critical
>
> It affects Storm releases which are greater than 0.9.3 so I didn't set affect version/s. All version lines are affected.
> When nextTuple is not calling at any chances (max spout pending, backpressure introduced at 1.0.0, and so on) multi-lang spout don't receive any message from ShellSpout, thus no messages from multi-lang spout side. (because multilang spout and ShellSpout communicates like ping-pong, unlike multilang bolt and ShellBolt)
> In result, no heartbeat is being marked, and finally raises subprocess heartbeat timeout.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)