> On March 15, 2017, 5:48 p.m., Alejandro Fernandez wrote:
> > Can we do the same thing in all places where tarballs are copied.
> > I know Hive attempts to copy sqoop tarball, but it can potentially fail if 
> > Sqoop is not installed as a service.
> > 
> > Perhaps we should actually check the topology for # of say Tez/Sqoop/Slider 
> > clients as opposed to just the existence of the config.

hello Alejandro,

I refactored the code to do the check in copy_tarball.py instead of in each 
service. I think this centralized implementation makes it easier to maintain. 
The check is against if clients are installed on the host.

Please review the new patch.

Thanks.


- Di


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/57579/#review169034
-----------------------------------------------------------


On March 17, 2017, 12:53 p.m., Di Li wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/57579/
> -----------------------------------------------------------
> 
> (Updated March 17, 2017, 12:53 p.m.)
> 
> 
> Review request for Ambari and Alejandro Fernandez.
> 
> 
> Bugs: AMBARI-20400
>     https://issues.apache.org/jira/browse/AMBARI-20400
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> It's possible to install a cluster without Tez and Slider via blueprint. In 
> this case, Yarn historyserver start logic prints two warnings about unable to 
> copy tez and slider tar ball to HDFS.
> Yarn history server should skip the copy hdfs tar ball logic for tez and 
> slider if they are not installed.
> 
> 
> Diffs
> -----
> 
>   
> ambari-common/src/main/python/resource_management/libraries/functions/copy_tarball.py
>  31a9be4 
>   ambari-server/src/test/python/stacks/2.0.6/configs/default.json fa7419f 
> 
> 
> Diff: https://reviews.apache.org/r/57579/diff/2/
> 
> 
> Testing
> -------
> 
> unit tests
> 
> 
> Thanks,
> 
> Di Li
> 
>

Reply via email to