[
https://issues.apache.org/jira/browse/MAPREDUCE-5329?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13776736#comment-13776736
]
Siddharth Seth commented on MAPREDUCE-5329:
-------------------------------------------
bq. It is true that shuffle-consumer is at job level. However, here we are
dealing with shuffle-provider which is an aux-service that serves multiple
jobs. Hence, it is not at job level.
An AM is per job - and the config being used is really a per AM/job
configuration. The config is just saying that this particular job may use an
alternate shuffle mechanism - but isn't related to what services will be run on
the cluster. The NM would have this configured at a cluster level (equivalent
to the TT config).
bq. last comment, we already have shuffle.consumer.plugin in mapred-site.xml,
wouldn't it will be appropriate to have the shuffle.provider.plugin next to it.
This is not critical for me, just let me know what you prefer.
I prefer leaving it out. Also it'd be good to have a comment which says that
this needs to be able to work with the host:port information provided by the AM
(i.e. plugins which require custom location / other configuration are not
supported)
bq. I'll surround it with ...
Shouldn't an attempt be made to initialize plugins which are specified by the
job ? Walking over configured NM plugins ends up potentially skipping plugins
which the job may have configured, but are not running in the NodeManager.
> APPLICATION_INIT is never sent to AuxServices other than the builtin
> ShuffleHandler
> -----------------------------------------------------------------------------------
>
> Key: MAPREDUCE-5329
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5329
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mr-am
> Affects Versions: 2.1.0-beta, 2.0.6-alpha
> Reporter: Avner BenHanoch
> Fix For: trunk
>
> Attachments: MAPREDUCE-5329.patch
>
>
> APPLICATION_INIT is never sent to AuxServices other than the built-in
> ShuffleHandler. This means that 3rd party ShuffleProvider(s) will not be
> able to function, because APPLICATION_INIT enables the AuxiliaryService to
> map jobId->userId. This is needed for properly finding the MOFs of a job per
> reducers' requests.
> NOTE: The built-in ShuffleHandler does get APPLICATION_INIT events due to
> hard-coded expression in hadoop code. The current TaskAttemptImpl.java code
> explicitly call: serviceData.put (ShuffleHandler.MAPREDUCE_SHUFFLE_SERVICEID,
> ...) and ignores any additional AuxiliaryService. As a result, only the
> built-in ShuffleHandler will get APPLICATION_INIT events. Any 3rd party
> AuxillaryService will never get APPLICATION_INIT events.
> I think a solution can be in one of two ways:
> 1. Change TaskAttemptImpl.java to loop on all Auxiliary Services and register
> each of them, by calling serviceData.put (…) in loop.
> 2. Change AuxServices.java similar to the fix in: MAPREDUCE-2668
> "APPLICATION_STOP is never sent to AuxServices". This means that in case the
> 'handle' method gets APPLICATION_INIT event it will demultiplex it to all Aux
> Services regardless of the value in event.getServiceID().
> I prefer the 2nd solution. I am welcoming any ideas. I can provide the
> needed patch for any option that people like.
> See [Pluggable Shuffle in Hadoop
> documentation|http://hadoop.apache.org/docs/current/hadoop-mapreduce-client/hadoop-mapreduce-client-core/PluggableShuffleAndPluggableSort.html]
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
