| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: only 1 task running instead of 4 - why ??? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
erich56
Senior Cruncher Austria Joined: Feb 24, 2007 Post Count: 300 Status: Offline Project Badges: 
|
I have been WCG cruncher for long time, on many PCs throughout many years.
Recently, I got a new PC with a 6-core HT CPU, i.e. 12 cores altogether. I am running 2 GPUGRID tasks (using my 2 high-end GPUs), 4 ATLAS@home tasks, and now I had planned to add 4 tasks for WCG - in order to have 10 tasks running altogether. The ressource share for GPUGRID is at 1000 (so that two such tasks are running in any case), the ressource share for ATLAS presently is at 200, the ressource share for WCG presently is at even 300. In the BOINC computing preferences, CPU usage is at 80%, which means 10 tasks (for 12 cores). When I added the WCG project, 2 GPUGRID tasks and 4 ATLAS tasks were running (for other reasons, by a config_app.xml I had limited ATLAS to maximum 4 tasks), so I expected WCG to download and run 4 tasks. However, only 1 task is running. When I go to "projects" and "update", the BOINC event log says: 17/04/2016 09:27:36 | World Community Grid | update requested by user 17/04/2016 09:27:39 | World Community Grid | Sending scheduler request: Requested by user. 17/04/2016 09:27:39 | World Community Grid | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: ) 17/04/2016 09:27:41 | World Community Grid | Scheduler request completed why so ??? CPU job cache is definitely NOT full with 7 tasks running - as said, before I had 10 task running with CPU usage set to 80% of my 12-core CPU. Any comments / help ? I would definitely like to have 4 WCG task running, not only 1. |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges: 
|
You didn't mention how many ATLAS tasks you have 'Ready to start'.
By limiting the number of ATLAS-tasks in progress, that doesn't mean 'Ready to start'-tasks doesn't count for the seconds you have in cache. And as noted in your messages the CPU-cache is full. |
||
|
|
erich56
Senior Cruncher Austria Joined: Feb 24, 2007 Post Count: 300 Status: Offline Project Badges:
|
yes, you are right - there were at least 4 or 5 ATLAS tasks "ready to start". Okay, I was not aware that nun-running tasks would also count in this sense.
----------------------------------------However, I realized this right away when I put the ATLAS project on "suspend" - immediately several WCG tasks were downloaded and started. In order to have 4 ea. ATLAS and WCG tasks running, I resumed the ATLAS project and aborted 4 WCG tasks; and indeed, then (beside the 2 GPUGRID tasks) 4 ea. WCG and ATLAS were running. Before I did all this, I set the ressource share of WCG to the same value as ATLAS (300). GPUGRID is at 1000. I then went away, and after I came back some 2 hours later, 8 (!) WCG tasks were running (beside several more on "ready to start"), and all 4 ATLAS tasks, that were runnig before, were on "waiting to run". How come? With same ressource share, one would expect that the number of running tasks would be about equal, or do I miss something? Would it help to setup a similar config_app.xml as I have it for ATLAS, limiting the number of runnings tasks to 4 ? What do you think? What would be the wording? The ATLAS is like this: <app_config> <app> <name>ATLAS</name> <max_concurrent>4</max_concurrent> </app> </app_config> [Edit 1 times, last edit by erich56 at Apr 17, 2016 12:28:31 PM] |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges:
|
The resource share is for the long term not for the actual running tasks.
----------------------------------------Long term could even mean what projects have run and how long over the last 30 days. It's hard to manipulate with several projects. The resource share for GPUGRID doesn't matter, cause I suppose that's the only GPU-project for you to run and you don't get CPU-tasks from GPUGRID. To start with an equal share you have to stop BOINC and edit the client_state.xml and change the value between <rec>X.XXXXXX</rec> into 0.000000 for all projects. The simplest app_config.xml for WCG (no matter what WCG-subprojects you are running) would be: <app_config> <project_max_concurrent>4</project_max_concurrent> </app_config> [Edit 1 times, last edit by Crystal Pellet at Apr 17, 2016 4:13:08 PM] |
||
|
|
erich56
Senior Cruncher Austria Joined: Feb 24, 2007 Post Count: 300 Status: Offline Project Badges:
|
The simplest app_config.xml for WCG (no matter what WCG-subprojects you are running) would be: <app_config> <project_max_concurrent>4</project_max_concurrent> </app_config> thank you very much; it's as simple as that  Works fine ! |
||
|
|
|