[
https://issues.apache.org/jira/browse/MAPREDUCE-2477?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Bill Graham updated MAPREDUCE-2477:
-----------------------------------
Description:
Initially opened against Pig, but this seems to be a MR bug.
When submitting a [MapReduce] job, the client uses the {{fs.default.name}}
supplied to it by the JobTracker (via core-site.xml on the master typically)
during the staging phase. After that, the client then uses the
{{fs.default.name}} from it's local configs. This seems like a bug to me.
Expected behavior would be to always use the local value.
I found this bug when the server configs were set to not use a FQDN for
{{fs.default.name}}. This caused the client to fail because it didn't have the
same default DNS domain.
was:
When submitting a Pig job, the client uses the {{fs.default.name}} supplied to
it by the JobTracker (via core-site.xml on the master typically) during the
staging phase. After that, the client then uses the {{fs.default.name}} from
it's local configs. This seems like a bug to me. Expected behavior would be to
always use the local value.
I found this bug when the server configs were set to not use a FQDN for
{{fs.default.name}}. This caused the client to fail because it didn't have the
same default DNS domain.
Summary: MR client uses fs.default.name as provided from the JobTracker
instead of local value (was: Pig client uses fs.default.name as provided from
the JobTracker instead of local value)
> MR client uses fs.default.name as provided from the JobTracker instead of
> local value
> -------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-2477
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2477
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Bill Graham
> Attachments: pig_1304465896181.log
>
>
> Initially opened against Pig, but this seems to be a MR bug.
> When submitting a [MapReduce] job, the client uses the {{fs.default.name}}
> supplied to it by the JobTracker (via core-site.xml on the master typically)
> during the staging phase. After that, the client then uses the
> {{fs.default.name}} from it's local configs. This seems like a bug to me.
> Expected behavior would be to always use the local value.
> I found this bug when the server configs were set to not use a FQDN for
> {{fs.default.name}}. This caused the client to fail because it didn't have
> the same default DNS domain.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
