[ 
http://opencast.jira.com/browse/MH-7676?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=26441#comment-26441
 ] 

Ruediger Rolf commented on MH-7676:
-----------------------------------

Hi Josh, 

thanks for the hints. I added the max load and will see if this brings a better 
load balancing. You may be right that vm161 has 4 cores activated for thsi VM 
(this was tested at the beginning of the last semester, so I nearly forgot 
this). But for load balancing I would prefer it if all machines are getting a 
job first and than I would assign additional jobs to the machines that have a 
higher load. Currrently the second encoder was only ideling and the first was 
stuck in too many jobs. The bottleneck for the performance according to the 
people in our IT-center is not the processing power but the disk IO (I did not 
veri this) but I would expect that maybe it would get faster if I make sure 
that the jobs are distributed more equal on all workers.


I activated debug mode and still encountered  the phenomenon that the worker 
start encoding again, altough the workflow service does not seem to know about 
this (at least from the info level logs at the admin server):
2011-05-09 10:15:07  INFO (CaptureAgentStateServiceImpl:423) - Recording status 
changed to 'capturing', resuming workflow '196'

Here is an excerpt of the lengthy debug logs on one of the restarted workers. I 
previously deleted the /tmp dir in the kope that this may cleanup the queued 
jobs:
http://pastebin.com/eYKkVZaL

I get an 404 error if I open /workflow/instance/892.xml

> Jobs (on distributed worker) sometimes only start after the felix has been 
> restarted.
> -------------------------------------------------------------------------------------
>
>                 Key: MH-7676
>                 URL: http://opencast.jira.com/browse/MH-7676
>             Project: Matterhorn Project
>          Issue Type: Bug
>          Components: Architecture & Services
>    Affects Versions: 1.1 
>            Reporter: Ruediger Rolf
>


--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________

Reply via email to