oodt-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christian Alan Mattmann <mattm...@usc.edu>
Subject Re: Problem with OODT Workflow RSUBMIT
Date Mon, 10 Nov 2014 03:38:14 GMT
Hi Zichuan,

-----Original Message-----

From: Zichuan Wang <zichuanw@usc.edu>
Date: Sunday, November 9, 2014 at 6:31 PM
To: Chris Mattmann <mattmann@usc.edu>
Subject: Re: Problem with OODT Workflow RSUBMIT

>Dear Professor,
>
>
>Thank you so much for replying.
>
>
> I’ve already set the max stack size of resmgr to 10000000, but there is
>no queue size that I can configure in the resource.properties file.

Gotcha, yea that is what I was talking about (the stack size).
Another property to set is the speed at which the scheduler wakes up
and picks a job off the queue.

> 
>
>
>And for batch_stab nodes, I’ve already started 10 of them, each with 1000
>capacity. I noticed that there is a jobs folder in resmgr’s policy
>folder, do I need to configure jobs before sending jobs from workflow
>manager?

Nope. If you set 10 batch stubs, with 1000 capacity, did you run them each
on a 
different port and make sure to configure them correctly in your nodes.xml
and
node-to-queue-map.xml file? What do you see in the resmgr log? Is it
trying to
allocate the jobs to batch stubs and is that part working?

>
>
>Also I’m using PGE 0.3, is it possible that’s the reason for the low
>speed?

Probably not - it’s a workflow/resource manager issue - see my questions
above and we can figure it out.

Cheers,
Chris

>
>
>Looking forward to your reply.
>Best regards,
>
>—
>Zichuan Wang
>Department of Computer Science, USC

++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Chris Mattmann, Ph.D.
Adjunct Associate Professor, Computer Science Department
University of Southern California
Los Angeles, CA 90089 USA
Email: mattmann@usc.edu
WWW: http://sunset.usc.edu/~mattmann/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++



>
>
>On Sun, Nov 9, 2014 at 9:46 AM, Christian Alan Mattmann
><mattmann@usc.edu> wrote:
>
>
>If it¹s talking to the resource manager my guess is it¹s
>either (a) not pushing through there b/c of the queue size;
>and or (b) the capacity on the batch stub.
>
>Try upping the batch stub capacity in resmgr/policy/nodes.xml.
>Maybe make it equal to the max # of jobs you want it running
>at a time? 
>
>Cheers, 
>Chris 
>
>++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>Chris Mattmann, Ph.D.
>Adjunct Associate Professor, Computer Science Department
>University of Southern California
>Los Angeles, CA 90089 USA
>Email: mattmann@usc.edu
>WWW: http://sunset.usc.edu/~mattmann/
>++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>
>
>
>
>-----Original Message-----
>From: Zichuan Wang <zichuanw@usc.edu>
>Date: Sunday, November 9, 2014 at 7:02 AM
>To: Chris Mattmann <mattmann@usc.edu>
>Subject: Re: Problem with OODT Workflow RSUBMIT
>
>> 
>> 
>> 
>> 
>>Here is a screenshot of OODT monitor. My wild guess is workflow is
>>running sequentially, instead of parallelly. Really got no idea what I
>>could do to make it faster...
>> 
>>‹ 
>>Zichuan Wang 
>>Department of Computer Science, USC
>> 
>> 
>>On Sun, Nov 9, 2014 at 5:06 AM, Zichuan Wang <zichuanw@usc.edu> wrote:
>> 
>> 
>>Dear Professor, 
>> 
>> 
>>I followed your advise on enlarge workflow manager¹s min pool size, and
>>now jobs no longer hang in queues, instead, they are marked as
>>³RSUBMIT². 
>>Could you please give me another advise on how to speed workflow manager
>>in this case? 
>> 
>>‹ 
>>Zichuan Wang 
>>Department of Computer Science, USC
>> 
>> 
>> 
>> 
>
>
>
>
>
>

Mime
View raw message