Message boards : ATLAS application : Small number of test tasks
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28617 - Posted: 22 Jan 2017, 21:30:17 UTC

There are 4 unsent WUs for Atlas:

22/01/2017 22:22:37 | LHC@home | Sending scheduler request: To fetch work.
22/01/2017 22:22:37 | LHC@home | Requesting new tasks for CPU
22/01/2017 22:22:39 | LHC@home | Scheduler request completed: got 0 new tasks
22/01/2017 22:22:39 | LHC@home | No tasks sent
22/01/2017 22:22:39 | LHC@home | No tasks are available for ATLAS Simulation



Supporting BOINC, a great concept !
ID: 28617 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28632 - Posted: 23 Jan 2017, 14:14:57 UTC

Now there are more than 100 Atlas-tasks, but still I get:

23/01/2017 15:11:50 | LHC@home | Sending scheduler request: To fetch work.
23/01/2017 15:11:50 | LHC@home | Requesting new tasks for CPU
23/01/2017 15:11:52 | LHC@home | Scheduler request completed: got 0 new tasks
23/01/2017 15:11:52 | LHC@home | No tasks sent
23/01/2017 15:11:52 | LHC@home | No tasks are available for ATLAS Simulation


Supporting BOINC, a great concept !
ID: 28632 · Report as offensive     Reply Quote
David Cameron
Project administrator
Project developer
Project scientist

Send message
Joined: 13 May 14
Posts: 387
Credit: 15,314,184
RAC: 0
Message 28633 - Posted: 23 Jan 2017, 15:08:49 UTC - in response to Message 28632.  

Same for me, and I saw others complaining about getting WU: https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4101

I've asked the admins to take a look.
ID: 28633 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28639 - Posted: 24 Jan 2017, 14:04:43 UTC

Got 4 Atlas-WUs

They seem to be 1-Core

So far I didn't set up an app_config, I wanted to see what happens without.

My LHC-Project-Settings:

Max # Jobs: No Limt
Max # CPUs: 4


Supporting BOINC, a great concept !
ID: 28639 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28641 - Posted: 24 Jan 2017, 15:42:14 UTC - in response to Message 28542.  

Will the Atlas-WUs here take into account the settings regarding NumberOfWUs and NumberOfCores from the Website or do we have to setup an app_config ?


They should do as they act upon the scheduling in the server rather than the client. I am just testing now for myself but there are no tasks.

Edit: It doesn't work at the moment but we should be able to get it working.

Could you make some progress on this ?

Is "MAX # CPUs" the number for the cores in an WU, so when I set this to 4, I get 4-Core-WUs for Atlas ?


Supporting BOINC, a great concept !
ID: 28641 · Report as offensive     Reply Quote
David Cameron
Project administrator
Project developer
Project scientist

Send message
Joined: 13 May 14
Posts: 387
Credit: 15,314,184
RAC: 0
Message 28643 - Posted: 24 Jan 2017, 15:59:44 UTC

Due to the large number of sixtrack tasks in the queue it's very hard to get an ATLAS (or any other app) task at the moment.

In my config I have

Max # jobs No limit
Max # CPUs 2

and 4-cores on my PC.

When I get ATLAS WU I get 2 at a time which makes sense, but they run single-core only.

Has anyone tried app_config.xml?
ID: 28643 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28644 - Posted: 24 Jan 2017, 16:01:31 UTC - in response to Message 28643.  

David wrote:
Has anyone tried app_config.xml?

I'm just working on it, but wanted to let the 4x 1-Core-WUs first finish or should I abort them?


Supporting BOINC, a great concept !
ID: 28644 · Report as offensive     Reply Quote
Crystal Pellet
Volunteer moderator
Volunteer tester

Send message
Joined: 14 Jan 10
Posts: 1268
Credit: 8,421,616
RAC: 2,139
Message 28645 - Posted: 24 Jan 2017, 17:32:30 UTC - in response to Message 28643.  
Last modified: 24 Jan 2017, 17:35:58 UTC

Has anyone tried app_config.xml?

Finally got an ATLAS-task. I already had the app_config prepared to run a 4 core VM.

It just started 5 minutes ago. When it's 1 job using 4 cores, it will run for about 2.5 hours, I think.

The single core I had before used 9 hours and 40 minutes (elapsed)
ID: 28645 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28647 - Posted: 24 Jan 2017, 18:30:12 UTC - in response to Message 28644.  

David wrote:
Has anyone tried app_config.xml?

I'm just working on it, but wanted to let the 4x 1-Core-WUs first finish or should I abort them?

Just downloaded a new WU with app_config, it is a 4-Core-WU!

For safety reasons I had changed the Web-Config "Max # CPU" and "Max # Jobs" to <No limit>


Supporting BOINC, a great concept !
ID: 28647 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,670,849
RAC: 235,955
Message 28648 - Posted: 24 Jan 2017, 19:52:22 UTC
Last modified: 24 Jan 2017, 20:03:31 UTC

I just got one with the setting no limit, I now have one atlas using 24cores.

That's not what I expected!

Setting my cores preference to 2 did what I expected.

did this ever get fixed?

http://atlasathome.cern.ch/forum_thread.php?id=534&postid=4977#4977
ID: 28648 · Report as offensive     Reply Quote
Crystal Pellet
Volunteer moderator
Volunteer tester

Send message
Joined: 14 Jan 10
Posts: 1268
Credit: 8,421,616
RAC: 2,139
Message 28649 - Posted: 24 Jan 2017, 20:19:33 UTC - in response to Message 28645.  

It just started 5 minutes ago. When it's 1 job using 4 cores, it will run for about 2.5 hours, I think.

The 4-core VM (configured with app_config) has finished as expected =>> https://lhcathome.cern.ch/lhcathome/result.php?resultid=115673716
ID: 28649 · Report as offensive     Reply Quote
David Cameron
Project administrator
Project developer
Project scientist

Send message
Joined: 13 May 14
Posts: 387
Credit: 15,314,184
RAC: 0
Message 28650 - Posted: 24 Jan 2017, 20:46:08 UTC - in response to Message 28648.  

I just got one with the setting no limit, I now have one atlas using 24cores.

That's not what I expected!


I'd be interested to know if that WU works and its performance (bad I expect)! We need to limit to max 8 cores.

did this ever get fixed?

http://atlasathome.cern.ch/forum_thread.php?id=534&postid=4977#4977


No, but I think the project preferences settings that Laurence introduced should avoid the problem because the server knows when it sends the WU the number of cores and how much memory it will use.
ID: 28650 · Report as offensive     Reply Quote
Claus Varming Lund

Send message
Joined: 15 Feb 15
Posts: 7
Credit: 26,671,002
RAC: 0
Message 28652 - Posted: 24 Jan 2017, 21:37:17 UTC

I've got some 28 Core WU's, after 90 sec they are only using 50% of one core - 3,8% of total CPU,

So limiting to 8 core might be a good idea.
ID: 28652 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,670,849
RAC: 235,955
Message 28658 - Posted: 25 Jan 2017, 7:13:18 UTC - in response to Message 28650.  

I quit it but I can change it back to try again and let you know. It will be on my 14core/28thread computer so it's contained within one CPU.
ID: 28658 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,369,412
RAC: 10,065
Message 28692 - Posted: 27 Jan 2017, 10:34:04 UTC

Nils Høimyr wrote:
We have restarted our daemons with the --allapps flag for the feeder now.

Please let us know if you get CMS and Theory tasks quicker now.

27/01/2017 11:30:13 | LHC@home | Sending scheduler request: To fetch work.
27/01/2017 11:30:13 | LHC@home | Requesting new tasks for CPU
27/01/2017 11:30:14 | LHC@home | Scheduler request completed: got 0 new tasks
27/01/2017 11:30:14 | LHC@home | No tasks sent
27/01/2017 11:30:14 | LHC@home | No tasks are available for ATLAS Simulation

Atlas shows 10 WUs to send on ServerStatusPage, so is Atlas out of WUs or does the scheduler still have a problem ?


Supporting BOINC, a great concept !
ID: 28692 · Report as offensive     Reply Quote
David Cameron
Project administrator
Project developer
Project scientist

Send message
Joined: 13 May 14
Posts: 387
Credit: 15,314,184
RAC: 0
Message 28694 - Posted: 27 Jan 2017, 14:26:07 UTC - in response to Message 28692.  

I've had no ATLAS tasks for 2 days now, but the LHC@Home guys are working on a solution to the scheduling problem.

I have put a few more WU in the queue in case people get lucky enough to have an ATLAS task scheduled.
ID: 28694 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,670,849
RAC: 235,955
Message 28705 - Posted: 29 Jan 2017, 7:49:16 UTC

I tried to do 2 24core tasks but they both seemed to turn into the never ending tasks.

I quit one after 9hrs and the other after 22hrs.

I can try again if there is some help to debug why they hung?
ID: 28705 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2071
Credit: 156,083,793
RAC: 105,610
Message 28708 - Posted: 29 Jan 2017, 8:44:50 UTC - in response to Message 28650.  

I'd be interested to know if that WU works and its performance (bad I expect)! We need to limit to max 8 cores.


+1 David
ID: 28708 · Report as offensive     Reply Quote
gyllic

Send message
Joined: 9 Dec 14
Posts: 202
Credit: 2,533,875
RAC: 0
Message 28754 - Posted: 31 Jan 2017, 20:57:12 UTC

i have crunched 4 4core multicore atlas tasks today. they all had pretty much the same running time and cpu time.
for three of them i got about 64 credits, but for the last one i got about 450.
so i wondered why i get nearly 7 times more points for pretty much the same computation time.
does anyone else have seen this behaviour? have you changed something or is this a bug?

here are the wus im talking about:
https://lhcathome.cern.ch/lhcathome/result.php?resultid=117806105
https://lhcathome.cern.ch/lhcathome/result.php?resultid=117805627
https://lhcathome.cern.ch/lhcathome/result.php?resultid=117801849
https://lhcathome.cern.ch/lhcathome/result.php?resultid=117678817
ID: 28754 · Report as offensive     Reply Quote
Tom*

Send message
Joined: 11 Aug 11
Posts: 6
Credit: 16,714,519
RAC: 0
Message 28756 - Posted: 1 Feb 2017, 0:24:36 UTC - in response to Message 28754.  

Just cross your fingers that creditnew (screw) keeps the larger credit value for
your ATLAS tasks.

It usually takes 11 validated tasks for creditnew to determine a baseline for your system, Since you switched from sixtrack to ATLAS with a bunch of invalidated sixtrack tasks it will be interesting to see what creditnew does.

from this point forward.
ID: 28756 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : ATLAS application : Small number of test tasks


©2024 CERN