[jira] [Commented] (STORM-266) Improve debuggability of ShellBolt and ShellSpout

["Parth Brahmbhatt (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/STORM-266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14034412#comment-14034412 ] 

Parth Brahmbhatt commented on STORM-266:
----------------------------------------

I have a patch  ready for this issue but I wanted to ask 2 questions before I submit for review:

1) When there is an exception , what is the purpose for waiting for the exit code? Based on the code it seems either by the time ShellProcess.java sees an error either the child process has exited or it has no knowledge of failure so it will continue running. In both cases waiting will not be helpful. 
2) As a general best practice I have observed in code we either log a message or we put the message as part of exception but never both. This item seems to be asking for both. Any reason?


> Improve debuggability of ShellBolt and ShellSpout
> -------------------------------------------------
>
>                 Key: STORM-266
>                 URL: https://issues.apache.org/jira/browse/STORM-266
>             Project: Apache Storm (Incubating)
>          Issue Type: Improvement
>            Reporter: Robert Joseph Evans
>
> ShellBolt and ShellSpout are sometimes difficult to debug, especially when there are multiple instances running in under the same worker.  It would be good to include the name of the bolt/spout in the logged messages, and probably the PID of the sub process too.
> In the case of errors it would be good for ShellProcess to try
> and get the exit code for the process, up to a timeout, and then get the
> contents of stderr.  Also in the case of errors when wrapping the exceptions we probably want to include the name of the bolt/spout and the PID in the
> Exception message.



--
This message was sent by Atlassian JIRA
(v6.2#6252)