Message boards : CMS Application : Large jobs?
Message board moderation

To post messages, you must log in.

AuthorMessage
Charles Huber

Send message
Joined: 26 Jun 16
Posts: 1
Credit: 1,505,456
RAC: 0
Message 28025 - Posted: 30 Nov 2016, 22:50:56 UTC
Last modified: 30 Nov 2016, 23:42:32 UTC

Are large, vLHC-style CMS jobs still a thing? I.e.:
...
Guest Log: [INFO] Reading volunteer information
Guest Log: [INFO] Volunteer: Charles Huber (437632) Host: 10408527
Guest Log: [INFO] Hey! You're from Kansas City.
Guest Log: [INFO] We have something special for you.
Guest Log: [INFO] VMID: 6d0ae20b-f23e-4d5d-b5ca-600a8fb1d26c
...


I spot-checked a few recent completed WUs and didn't see any like that.
ID: 28025 · Report as offensive     Reply Quote
ivan
Volunteer moderator
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar

Send message
Joined: 29 Aug 05
Posts: 998
Credit: 6,264,307
RAC: 71
Message 28029 - Posted: 1 Dec 2016, 9:16:07 UTC - in response to Message 28025.  

I don't think we're doing that at the moment. Ben would know, but he's in Paris until Monday. :-)
ID: 28029 · Report as offensive     Reply Quote
Profile Laurence
Project administrator
Project developer

Send message
Joined: 20 Jun 14
Posts: 372
Credit: 238,712
RAC: 0
Message 28030 - Posted: 1 Dec 2016, 10:44:13 UTC - in response to Message 28029.  

I can confirm that we are not doing large jobs at the moment. We have everything is place but are going to focus on the project consolidation, stability and reliability before we attempt anything more ambitious.
ID: 28030 · Report as offensive     Reply Quote
Stick

Send message
Joined: 21 Aug 07
Posts: 46
Credit: 1,503,630
RAC: 0
Message 28045 - Posted: 2 Dec 2016, 23:57:24 UTC - in response to Message 28030.  

My CMS jobs are all having problems and the symptoms are indicative of being too big for my computers. That is, all were suspended (waiting on memory) until BOINC finished all the tasks from other projects. Those jobs running on Computer 10342612 all finished and validated once the cache was cleared but BOINC could only run one CMS at a time and the computer was clearly slower responding when the CMS tasks were running. And the task which was assigned to Computer 9926211 finally failed with STATUS_STACK_BUFFER_OVERRUN after BOINC had cleared all other tasks from its cache and CMS started running.

I hope this is not indicative of how normal-size CMS jobs are supposed to perform. If so, I will have to opt out of running CMS application tasks.
ID: 28045 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2386
Credit: 222,955,082
RAC: 136,930
Message 28046 - Posted: 3 Dec 2016, 9:01:15 UTC - in response to Message 28045.  

CMS requests at least 2.33 GB of free RAM (singlecore; 2 GB for the VM).
Only 2 of your hosts have enough physical RAM to fulfill this requirement:
https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=10342612 (3981.84 MB; should be enough)
https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=9926211 (2662.86 MB; only slightly above; may be very sluggish due to heavy disk usage)
ID: 28046 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,770,056
RAC: 231,829
Message 28047 - Posted: 3 Dec 2016, 9:06:19 UTC

The Theory tasks are in good supply an only need 600MB of ram.
ID: 28047 · Report as offensive     Reply Quote
ivan
Volunteer moderator
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar

Send message
Joined: 29 Aug 05
Posts: 998
Credit: 6,264,307
RAC: 71
Message 28048 - Posted: 3 Dec 2016, 12:30:02 UTC - in response to Message 28045.  

My CMS jobs are all having problems and the symptoms are indicative of being too big for my computers. That is, all were suspended (waiting on memory) until BOINC finished all the tasks from other projects. Those jobs running on Computer 10342612 all finished and validated once the cache was cleared but BOINC could only run one CMS at a time and the computer was clearly slower responding when the CMS tasks were running. And the task which was assigned to Computer 9926211 finally failed with STATUS_STACK_BUFFER_OVERRUN after BOINC had cleared all other tasks from its cache and CMS started running.

I hope this is not indicative of how normal-size CMS jobs are supposed to perform. If so, I will have to opt out of running CMS application tasks.

We'll be sorry to see you go, but thank you for trying. We can't guarantee that our design decisions will be suitable for everyone, unfortunately, but your experience is valuable feedback.
ID: 28048 · Report as offensive     Reply Quote
Stick

Send message
Joined: 21 Aug 07
Posts: 46
Credit: 1,503,630
RAC: 0
Message 28051 - Posted: 3 Dec 2016, 19:39:52 UTC - in response to Message 28048.  
Last modified: 3 Dec 2016, 19:41:32 UTC

Thanks to all who replied. Obviously, I haven't fully absorbed the magnitude of all the changes that are coming via the LHC@home consolidation. It's a huge difference from when it was "only Sixtrack" and I have to admit that, until now, I haven't been paying close attention. I promise to do better. And, even though CMS may not be suitable for the computers I currently own, it might be the justification I need to upgrade. ;-)
ID: 28051 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,770,056
RAC: 231,829
Message 28052 - Posted: 4 Dec 2016, 10:43:30 UTC - in response to Message 28051.  

for your single core systems, the upgrade to 4Gb of ram shouldn't be too hard. The Pentium system can take up to 16Gbthis costs about $60
ID: 28052 · Report as offensive     Reply Quote

Message boards : CMS Application : Large jobs?


©2024 CERN