> On Nov. 3, 2014, 8:53 p.m., Jonathan Hurley wrote:
> > What happens when this agent code is installed in a cluster that's not 
> > running HDFS (GlusterFS for example)?

Presently, the command will fail and will not push the log files to HDFS.  For 
log to work, we need HDFS running.


The LogHandlerCommand on the ambari-server side can skip pushing the configs if 
HDFS service is not available.  In that case, no jobs will be scheduled.


> On Nov. 3, 2014, 8:53 p.m., Jonathan Hurley wrote:
> > ambari-agent/src/main/python/ambari_agent/loghandler/hdfsapi.py, line 78
> > <https://reviews.apache.org/r/27396/diff/2-3/?file=746522#file746522line78>
> >
> >     command_args (spelling)

Fixed in the latest patch.


- Cabir


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27396/#review59638
-----------------------------------------------------------


On Nov. 4, 2014, 10:11 p.m., Cabir Zounaidou wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27396/
> -----------------------------------------------------------
> 
> (Updated Nov. 4, 2014, 10:11 p.m.)
> 
> 
> Review request for Ambari, Alejandro Fernandez, Jonathan Hurley, John 
> Speidel, Mahadev Konar, Nate Cole, and Yusaku Sako.
> 
> 
> Bugs: AMBARI-1522
>     https://issues.apache.org/jira/browse/AMBARI-1522
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> The idea is to run a scheduler which computes diffs in a component log and 
> pushes the diff (only if available) to HDFS at fixed interval.
> 
> 1. Each component will have its own monitor
> 2. The apscheduler threadpool controls the monitor execution.
> 3. If the log directory is not available if the component (or service) is not 
> running, it will skip the processing.
> 4. It saves the last read line index and last modified time for next iteration
> 5. It uses HDFS shell utility to push log patch to HDFS.
> 6. Right now, the component log directory is configured in a json file.  In 
> the next iteration, it will try to automatically detected from the stack 
> config.
> 
> 
> Diffs
> -----
> 
>   ambari-agent/src/main/python/ambari_agent/Controller.py dc3a1cf 
>   ambari-agent/src/main/python/ambari_agent/LogSchedulerHandler.py 
> PRE-CREATION 
>   ambari-agent/src/main/python/ambari_agent/loghandler/__init__.py 
> PRE-CREATION 
>   ambari-agent/src/main/python/ambari_agent/loghandler/config.py PRE-CREATION 
>   ambari-agent/src/main/python/ambari_agent/loghandler/hdfsapi.py 
> PRE-CREATION 
>   ambari-agent/src/main/python/ambari_agent/loghandler/monitor.py 
> PRE-CREATION 
>   ambari-agent/src/main/python/ambari_agent/loghandler/util.py PRE-CREATION 
>   ambari-agent/src/test/python/ambari_agent/TestLogHandler.py PRE-CREATION 
>   
> ambari-agent/src/test/python/ambari_agent/dummy_files/log_handler_config.json 
> PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/27396/diff/
> 
> 
> Testing
> -------
> 
> - Ran the ambari-agent with the patch and verified
> 1. Started ambari-agent with the patch 
> 2. The scheduler started successfully
> 3. The logs files are getting pushed to HDFS successfully. Verified using 
> HDFS shell utility.
> - Ran the tests successfully
> [INFO]
> [INFO] 
> ------------------------------------------------------------------------
> [INFO] Building Ambari Agent 1.3.0-SNAPSHOT
> [INFO] 
> ------------------------------------------------------------------------
> [INFO]
> :
> ----------------------------------------------------------------------
> Ran 324 tests in 8.282s
> 
> OK
> :
> [INFO] 
> ------------------------------------------------------------------------
> [INFO] BUILD SUCCESS
> [INFO] 
> ------------------------------------------------------------------------
> [INFO] Total time: 14.759s
> [INFO] Finished at: Thu Oct 30 14:58:21 PDT 2014
> [INFO] Final Memory: 10M/4079M
> [INFO] 
> ------------------------------------------------------------------------
> 
> 
> Thanks,
> 
> Cabir Zounaidou
> 
>

Reply via email to