[
https://issues.apache.org/jira/browse/HADOOP-4141?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12630071#action_12630071
]
Aaron Kimball commented on HADOOP-4141:
---------------------------------------
As a related comment:
The patch I've submitted will log a warning-level message when the resolver is
actually executed, and return an error value (null).
There's currently no great way to report these issues at configure()-time. You
can throw an unchecked exception if you'd like, but should we maybe think about
adding a checked MisconfigurationException so that if configure() tries to load
a value that isn't valid, we have a structured way of propagating these results
to the client?
> ScriptBasedMapping has infinite loop with invalid jobconf parameter
> -------------------------------------------------------------------
>
> Key: HADOOP-4141
> URL: https://issues.apache.org/jira/browse/HADOOP-4141
> Project: Hadoop Core
> Issue Type: Bug
> Affects Versions: 0.18.0
> Reporter: Aaron Kimball
> Priority: Minor
> Attachments: scriptMap.patch, scriptMap.patch.2,
> TestScriptBasedMapping.java
>
> Original Estimate: 0h
> Remaining Estimate: 0h
>
> A few bugs in ScriptBasedTopology:
> 1) topology.script.number.args is not validated; if this is <= 0, the
> runResolveCommand() method will run its outer loop an infinite number of
> times.
> 2) CachedDNSToSwitchMapping assumes that if a List is returned, it has the
> same cardinality as the number of names. ScriptBasedMapping should return
> 'null' instead of a mis-sized list on error
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
