[
https://issues.apache.org/jira/browse/STORM-467?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Kania updated STORM-467:
------------------------------
Description:
When issuing the storm jar command, the command will launch with some cached
version of the jar contents even if no jar file is now present at the location
specified on the command line. This should instead cause an error so that a
user is actually running what they think they are.
The second part of this is that some part of storm is caching topology classes
so that when debugging errors, old code is executed instead of the new version
of a class. I would argue that storm should attempt to destroy cached topology
classes if presented with a new version or when an active topology is
terminated.
was:
When issuing the storm jar command, the command will launch with some cached
version of the jar contents even if no jar file is now present at the specified
location on the command line. This should instead cause an error to that a user
is actually running what they think they are.
The second part of this is that some part of storm is caching topology classes
so that when debugging errors, old code is executed instead of the new version
of a class. I would argue that storm should attempt to destroy cached topology
classes if presented with a new version or when an active topology is
terminated.
> storm jar executes non-existant or old jar files and classes
> ------------------------------------------------------------
>
> Key: STORM-467
> URL: https://issues.apache.org/jira/browse/STORM-467
> Project: Apache Storm (Incubating)
> Issue Type: Bug
> Affects Versions: 0.9.2-incubating
> Environment: debian linux wheezy and squeeze with java 1.7.0_67
> Reporter: Jason Kania
>
> When issuing the storm jar command, the command will launch with some cached
> version of the jar contents even if no jar file is now present at the
> location specified on the command line. This should instead cause an error so
> that a user is actually running what they think they are.
> The second part of this is that some part of storm is caching topology
> classes so that when debugging errors, old code is executed instead of the
> new version of a class. I would argue that storm should attempt to destroy
> cached topology classes if presented with a new version or when an active
> topology is terminated.
--
This message was sent by Atlassian JIRA
(v6.2#6252)