Message boards : Sixtrack Application : Very long Tasks
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Profile ritterm
Avatar

Send message
Joined: 30 May 08
Posts: 93
Credit: 5,160,246
RAC: 0
Message 39248 - Posted: 3 Jul 2019, 17:32:17 UTC

ID: 39248 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1611
Credit: 94,703,558
RAC: 98,795
Message 39250 - Posted: 3 Jul 2019, 17:35:58 UTC - in response to Message 39248.  

ID: 39250 · Report as offensive     Reply Quote
Alessio Mereghetti
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 29 Feb 16
Posts: 157
Credit: 2,549,656
RAC: 3,068
Message 39260 - Posted: 4 Jul 2019, 7:30:28 UTC - in response to Message 39246.  


Aren't those long task more suited to GPU's?

Well, GPUs are suitable for calculations that can be run in parallel. That could be the case for SixTrack only if the number of particles is kept constant with time, which implies all stable particles. But once you start hitting unstable regions of the phase space, you should switch back to regular CPUs.
Knowing this point a-priori would imply also knowing the dynamic aperture, which is exactly what we are looking at...
Anyway, the effort to go to GPUs is on-going, but that requires deep re-factoring of the code. It is not in our top-priority, but it is steadily progressing - no clear ETA, though...


Whoa...

Workunit 117930579

Workunit 117930567

Workunit 117930673

Great job, many many thanks!
ID: 39260 · Report as offensive     Reply Quote
Profile ritterm
Avatar

Send message
Joined: 30 May 08
Posts: 93
Credit: 5,160,246
RAC: 0
Message 39280 - Posted: 4 Jul 2019, 18:37:10 UTC - in response to Message 39260.  
Last modified: 4 Jul 2019, 18:37:33 UTC

Whoa...

Workunit 117930579

Workunit 117930567

Workunit 117930673

Great job, many many thanks!

Always a pleasure to grind it out for my favorite project... :-)
ID: 39280 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 1072
Credit: 36,374,369
RAC: 5,010
Message 40107 - Posted: 10 Oct 2019, 3:55:36 UTC

This Sixtrack(avx)-248217116 is shown with a runtime of more than 2 days for the moment (10-Time greater than normal)
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=124082860
ID: 40107 · Report as offensive     Reply Quote
Profile MAGIC Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 977
Credit: 41,730,252
RAC: 19,856
Message 40109 - Posted: 10 Oct 2019, 7:02:41 UTC - in response to Message 40107.  

This Sixtrack(avx)-248217116 is shown with a runtime of more than 2 days for the moment (10-Time greater than normal)
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=124082860


Those are the workspace1_HEL_Qp Sixtrack tasks taking close to 55 hours and I had 2 that just finished as Validate state - Workunit error - check skipped

So after a few of those I decided against wasting 55 hours and just abort them when I see them after a reload of tasks.
( I have just one of those that only ran 16 hours but have the same Validate state)


https://lhcathome.cern.ch/lhcathome/result.php?resultid=247499200
https://lhcathome.cern.ch/lhcathome/result.php?resultid=247174576
https://lhcathome.cern.ch/lhcathome/result.php?resultid=247174715

This one was instead a Exit status 196 (0x000000C4) EXIT_DISK_LIMIT_EXCEEDED Run time 16 hours 45 min 54 sec
https://lhcathome.cern.ch/lhcathome/result.php?resultid=247556609
ID: 40109 · Report as offensive     Reply Quote
Alessio Mereghetti
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 29 Feb 16
Posts: 157
Credit: 2,549,656
RAC: 3,068
Message 40115 - Posted: 10 Oct 2019, 10:49:54 UTC - in response to Message 40109.  

The issue with those tasks has been addressed here:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4690

We had a problem with the setting up of a bunch of simulations with 1^7 turns, and a log file was eating up all the disk quota.
The user was not aware that sixtrack was producing that log file, and how to update the request.
Hence, I have decided to manually attribute credit to those tasks based on the CPU time via the --credit_from_runtime option of the BOINC validator.
I should have also given credit to your tasks, but I had a malformed query to the DB - now everything should be properly set, and you should see credit to those tasks.

Thanks a lot in advance,
Happy cruncing,
A.
ID: 40115 · Report as offensive     Reply Quote
Profile MAGIC Quantum Mechanic
Avatar

Send message
Joined: 24 Oct 04
Posts: 977
Credit: 41,730,252
RAC: 19,856
Message 40125 - Posted: 11 Oct 2019, 10:46:32 UTC - in response to Message 40115.  

The issue with those tasks has been addressed here:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4690

We had a problem with the setting up of a bunch of simulations with 1^7 turns, and a log file was eating up all the disk quota.
The user was not aware that sixtrack was producing that log file, and how to update the request.
Hence, I have decided to manually attribute credit to those tasks based on the CPU time via the --credit_from_runtime option of the BOINC validator.
I should have also given credit to your tasks, but I had a malformed query to the DB - now everything should be properly set, and you should see credit to those tasks.

Thanks a lot in advance,
Happy cruncing,
A.


Thanks Alessio,
Glad I stopped by here again just now to see that was taken care off.

I have no problem with running long tasks as long as they end up being used after they are finished.
(especially over 55 hours each)

- Samson
ID: 40125 · Report as offensive     Reply Quote
Erich56

Send message
Joined: 18 Dec 15
Posts: 1322
Credit: 24,379,850
RAC: 10,495
Message 43812 - Posted: 10 Dec 2020, 11:51:07 UTC

A few days ago, when there were Sixtrack tasks available for a short period of time, on one of my PCs I downloaded quite a a number of them, because the first few ones got ready within seconds.
The other ones I downloaded turn out to be longrunners, it takes them about 10 hours to get finished (workspace2_hl14_collision_ ...). They all have deadline December 13. For sure, not all of them will get finished before that date.
Question for the experts: what should I do with the tasks with are left in the queue on December 13? Would it make sense to get them crunched, anyway, although they would upload beyond the deadline?
Or does it make best sense to delete them from the queue since there is no use in getting them crunched?
ID: 43812 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1611
Credit: 94,703,558
RAC: 98,795
Message 43813 - Posted: 10 Dec 2020, 12:01:16 UTC - in response to Message 43812.  

As soon as the deadline (+ a short grace period) is over the server marks the tasks as lost and initiates a resend.
Lost tasks don't get credits.
Hence, they should be cancelled.
ID: 43813 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 484
Credit: 25,918,455
RAC: 14,521
Message 43814 - Posted: 10 Dec 2020, 12:31:04 UTC - in response to Message 43813.  

Does Boinc cancel tasks that have not been started at deadline? I am using BoincTasks as a Boinc Manager substitute and that happens, but I don't know if that is a feature of Boinc or BoincTasks. Tasks that are already running at deadline will continue running until they are finished.
ID: 43814 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1611
Credit: 94,703,558
RAC: 98,795
Message 43863 - Posted: 11 Dec 2020, 20:57:07 UTC - in response to Message 43814.  

Does Boinc cancel tasks that have not been started at deadline?

The project server can cancel tasks at the client that are not yet started but I'm not sure if this is the default or just an option.

Tasks that are already running at deadline will continue running until they are finished.

Right, but they don't get credits if they are marked as overdue at the server.
ID: 43863 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 484
Credit: 25,918,455
RAC: 14,521
Message 43867 - Posted: 11 Dec 2020, 22:16:07 UTC - in response to Message 43863.  

Tasks that are already running at deadline will continue running until they are finished.

Right, but they don't get credits if they are marked as overdue at the server.

They might if you can return your result before the third host does.
ID: 43867 · Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : Sixtrack Application : Very long Tasks


©2021 CERN