hadoop-hdfs-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Anfernee Xu <anfernee...@gmail.com>
Subject Re: job submission between 2 YARN clusters
Date Fri, 14 Feb 2014 00:59:47 GMT
Hi Zhijie,

I agree, what I'm doing in the standalone app is that the app loads the
first cluster Configuration(mapred-site.xml, yarn-site.xml) as its default
configuration, and then submit MR job with this configuration to the first
cluster, and after the job is finished, I will submit the second job to the
second cluster with almost same Configuration exception I changed the
property: yarn.resourcemanager.address pointing to the second cluster's RM.
My guess the job.xml of the second job holds all property values of the
first cluster(such as yarn.resourcemanager.scheduler.address) and will
override these properties specified in the second cluster(yarn-site.xml for
example), therefore it will talk to the wrong RM when NM is launching the
container.

Please comment.

BTW, I just tweak the standalone app so that it will load the second
cluster's configuration(yarn-site.xml) before submit the second job, it
seems working.

Thanks


On Thu, Feb 13, 2014 at 4:28 PM, Zhijie Shen <zshen@hortonworks.com> wrote:

> Hi Anfernee,
>
> It sounds most likely that config somehow corrupts. So you have two sets
> of config to start two YARN cluster separately, don't you? If you provide
> more detail about how you config the two clusters, it's easy for the
> community to understand your problem.
>
> - Zhijie
>
>
> On Thu, Feb 13, 2014 at 11:34 AM, Anfernee Xu <anfernee.xu@gmail.com>wrote:
>
>> I'm at Yarn 2.2.0 release, I configured 2 single-node clusters on my
>> laptop(just for POC and all port conflicts are resolved, and I can see NM
>> and RM is up, webUI shows everything is fine) and I also have a standalone
>> java application. The java application is a kind of job client, it will
>> submit job1 to Cluser #1, once the job is finished, it will submit another
>> job2 to Cluster #2.
>>
>> What I'm seeing is the job1 is doing fine, but job2 failed, I looked
>> source code, and found the NM in cluser2 was talking to cluser1's RM via
>> wrong yarn.resourcemanager.scheduler.address. How that happens? I just want
>> to make sure there's no such issue in real deployment.
>>
>> --
>> --Anfernee
>>
>
>
>
> --
> Zhijie Shen
> Hortonworks Inc.
> http://hortonworks.com/
>
> CONFIDENTIALITY NOTICE
> NOTICE: This message is intended for the use of the individual or entity
> to which it is addressed and may contain information that is confidential,
> privileged and exempt from disclosure under applicable law. If the reader
> of this message is not the intended recipient, you are hereby notified that
> any printing, copying, dissemination, distribution, disclosure or
> forwarding of this communication is strictly prohibited. If you have
> received this communication in error, please contact the sender immediately
> and delete it from your system. Thank You.




-- 
--Anfernee

Mime
View raw message