Hi All,
  I submitted an application to YARN. But there are no containers
available. In this case job got hanged indefinitely.
  I reproduced this in a pseudo cluster mode with the following steps.

Steps to reproduce

1) let the amount of memory the MR app master needs be 2GB.
2) let the Nodemanager's capability be 1GB.

In the above case, no of available containers is 0 and so the job could
not able to get a chance to run due to containers unavailability.

Please validate the above case.

--
Thanks&Regards,
Bh.V.S.Kamesh.

Kamesh,

 This is what will happen if you do not have enough resource in the
cluster to run a job. Though there could be an improvement that a job
fails with not enough resource in the cluster but thats not what
happens right now.

thanks
mahadev

On Mon, Oct 17, 2011 at 6:25 AM, Kamesh <[EMAIL PROTECTED]> wrote:
> Hi All,
>  I submitted an application to YARN. But there are no containers available.
> In this case job got hanged indefinitely.
>  I reproduced this in a pseudo cluster mode with the following steps.
>
> Steps to reproduce
>
> 1) let the amount of memory the MR app master needs be 2GB.
> 2) let the Nodemanager's capability be 1GB.
>
> In the above case, no of available containers is 0 and so the job could not
> able to get a chance to run due to containers unavailability.
>
> Please validate the above case.
>
> --
> Thanks&Regards,
> Bh.V.S.Kamesh.
>
>

This sounds like
https://issues.apache.org/jira/browse/MAPREDUCE-2324manifested in MR2
to me. Should we perhaps update the "affects versions"
field there?

--
Aaron T. Myers
Software Engineer, Cloudera

On Mon, Oct 17, 2011 at 9:55 AM, Mahadev Konar <[EMAIL PROTECTED]>wrote:

> Kamesh,
>
>  This is what will happen if you do not have enough resource in the
> cluster to run a job. Though there could be an improvement that a job
> fails with not enough resource in the cluster but thats not what
> happens right now.
>
> thanks
> mahadev
>
> On Mon, Oct 17, 2011 at 6:25 AM, Kamesh <[EMAIL PROTECTED]> wrote:
> > Hi All,
> >  I submitted an application to YARN. But there are no containers
> available.
> > In this case job got hanged indefinitely.
> >  I reproduced this in a pseudo cluster mode with the following steps.
> >
> > Steps to reproduce
> >
> > 1) let the amount of memory the MR app master needs be 2GB.
> > 2) let the Nodemanager's capability be 1GB.
> >
> > In the above case, no of available containers is 0 and so the job could
> not
> > able to get a chance to run due to containers unavailability.
> >
> > Please validate the above case.
> >
> > --
> > Thanks&Regards,
> > Bh.V.S.Kamesh.
> >
> >
>

On Mon, Oct 17, 2011 at 11:08 AM, Aaron T. Myers <[EMAIL PROTECTED]> wrote:

> This sounds like https://issues.apache.org/jira/browse/MAPREDUCE-2324manifested in MR2 to me. Should we perhaps update the "affects versions"
> field there?
>

Never mind. I just found
MAPREDUCE-2723<https://issues.apache.org/jira/browse/MAPREDUCE-2723>
.

--
Aaron T. Myers
Software Engineer, Cloudera