log in

Deadline change for ATLAS jobs


Advanced search

Message boards : News : Deadline change for ATLAS jobs

Author Message
Wenjing Wu
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Project scientist
Send message
Joined: 23 Jun 14
Posts: 10
Credit: 16,547,346
RAC: 343,742
Message 31961 - Posted: 16 Aug 2017, 8:42:03 UTC

Due to the tight deadline of the ATLAS tasks, we change to deadline of ATLAS jobs from 2 weeks to 1 week. The ATLAS job takes about 3-4 CPU hours to finish on a moderate CPU (2.5GFLOPS).

Profile Yeti
Volunteer moderator
Avatar
Send message
Joined: 2 Sep 04
Posts: 281
Credit: 41,119,901
RAC: 51,084
Message 31962 - Posted: 16 Aug 2017, 12:58:25 UTC

essentially this seems to be okay, but:

All my clients, that got a task with 1-Week-Deadline are switching in panic-mode and start crunching the 1-Week-Deadline with "High-Priority". This indicates that something is wrong with this tasks. And as their are no other projects running on my clients, it can only be a problem with Atlas-Tasks themselve

Are the estimated flops set to a realistic number ?

The "High-Priority" and the complaints about varyiing credits ( https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4271 ) indicate that they are wrong
____________


Supporting BOINC, a great concept !

maeax
Send message
Joined: 2 May 07
Posts: 182
Credit: 11,314,906
RAC: 11,352
Message 31963 - Posted: 16 Aug 2017, 13:41:59 UTC

got the first two Atlas with this week-deadline.
In the LHCatHome Website is the deadline ONE day later as in the Boincmanager shown.

Crystal Pellet
Volunteer moderator
Volunteer tester
Send message
Joined: 14 Jan 10
Posts: 329
Credit: 2,779,814
RAC: 3,567
Message 31965 - Posted: 16 Aug 2017, 14:17:20 UTC - in response to Message 31962.

All my clients, that got a task with 1-Week-Deadline are switching in panic-mode and start crunching the 1-Week-Deadline with "High-Priority". This indicates that something is wrong with this tasks. And as their are no other projects running on my clients, it can only be a problem with Atlas-Tasks themselve

Maybe you've BOINC work buffer set to =>7 days?

tullio
Send message
Joined: 19 Feb 08
Posts: 420
Credit: 2,049,855
RAC: 336
Message 31971 - Posted: 16 Aug 2017, 16:56:53 UTC
Last modified: 16 Aug 2017, 16:57:57 UTC

I've started a one week task on my HP Linux laptop, with VirtualBox RAM equal to 4200 MB but it soon went into Waiting for memory. I changed the RAM to 5200 of a total of 8000, but it still goes Waiting for memory. There is only another task running on the HP, a climateprediction.net with a very extended deadline (one year). The CPU is an AMD E-450 with 1.6 GHz, two cores.
Tullio

Toby Broom
Volunteer moderator
Send message
Joined: 27 Sep 08
Posts: 359
Credit: 78,419,513
RAC: 112,883
Message 31972 - Posted: 16 Aug 2017, 17:27:36 UTC - in response to Message 31971.
Last modified: 16 Aug 2017, 17:28:15 UTC

Hi Tullio, the setting in the appconfig are not passed to BOINC for it's calculations of memory consumption.

You can adjust the memory reported to BOINC by using the Max # CPUs setting (in web prefernces), where 3 cores is reported as 4.88GB to boinc.

Since you use the app config to force the number of cores the websetting have no effect on the number of cores used.

tullio
Send message
Joined: 19 Feb 08
Posts: 420
Credit: 2,049,855
RAC: 336
Message 31973 - Posted: 16 Aug 2017, 18:24:53 UTC - in response to Message 31972.

Thanks Toby. No, I simply start the VirtualBox Manager and make my modifications in the Setting. But it seems to go on, even if rather jumpily, and I am letting it run.
Tullio

Profile rbpeake
Send message
Joined: 17 Sep 04
Posts: 55
Credit: 15,620,725
RAC: 1,342
Message 31974 - Posted: 16 Aug 2017, 18:46:54 UTC
Last modified: 16 Aug 2017, 18:50:05 UTC

There was discussion at some point of having a BOINC option to run longer work units that were being issued to the other computer centers. I guess there were some issues uncovered and it was never pursued further, but it worked well for me. It seemed like an efficient way to run.
____________
Regards,
Bob P.

computezrmle
Send message
Joined: 15 Jun 08
Posts: 347
Credit: 3,401,602
RAC: 3,491
Message 31975 - Posted: 16 Aug 2017, 19:32:20 UTC - in response to Message 31973.

tullio wrote:
I simply start the VirtualBox Manager and make my modifications in the Setting.

What do you modify directly with VirtualBox Manager?
To be honest I either completely misunderstand your method or I suspect this is part of the problems you described here:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4324&postid=31817
tullio wrote:
all LHC tasks except SixTrack fail on all 3 for unknown reasons.

Toby Broom
Volunteer moderator
Send message
Joined: 27 Sep 08
Posts: 359
Credit: 78,419,513
RAC: 112,883
Message 31981 - Posted: 17 Aug 2017, 10:41:12 UTC - in response to Message 31973.

This is why you see waiting for RAM in Boinc, it doesn't know about the changes you made.

If you want it to run 2 tasks the Boinc need to know about the system utilitisation

tullio
Send message
Joined: 19 Feb 08
Posts: 420
Credit: 2,049,855
RAC: 336
Message 31982 - Posted: 17 Aug 2017, 11:26:06 UTC - in response to Message 31975.

I run all stock applications and never modify their settings. ATLAS@home starts at 4200 MB and then goes in Waiting for memory. That is why I tried upgrading the memory to 5200 MB and now it is at 80% after 16 hours. LHC@home tasks fail on my WINDOWS 10 PC which has 24 GB RAM. I accept only SixTrack tasks on my 2 Linux boxes whose RAM tops at 8 GB. Now I am trying Atlas@home on the HP Laptop and it seems to run with RAM at 5200 MB.
Tullio

computezrmle
Send message
Joined: 15 Jun 08
Posts: 347
Credit: 3,401,602
RAC: 3,491
Message 31983 - Posted: 17 Aug 2017, 11:54:15 UTC - in response to Message 31982.

@Tullio

Here you wrote that you modified the settings:

tullio wrote:
I simply start the VirtualBox Manager and make my modifications in the Setting.

Here you wrote that you did not modify the settings:
tullio wrote:
I run all stock applications and never modify their settings.

That's confusing and makes it hard to help you.

Nonetheless, if you try to run ATLAS you may set the preferences on the project's website to "number of cores = 1" and rise the RAM settings via an app_config.xml that looks like:

<app_config>
<app>
<name>ATLAS</name>
<max_concurrent>1</max_concurrent>
</app>
<app_version>
<app_name>ATLAS</app_name>
<plan_class>vbox64_mt_mcore_atlas</plan_class>
<avg_ncpus>1.0</avg_ncpus>
<cmdline>--memory_size_mb 5000</cmdline>
</app_version>
</app_config>

Then cancel the currently running ATLAS WU, restart your computer and download a fresh WU to ensure the settings apply.

tullio
Send message
Joined: 19 Feb 08
Posts: 420
Credit: 2,049,855
RAC: 336
Message 31984 - Posted: 17 Aug 2017, 12:16:24 UTC - in response to Message 31983.

I think that all BOINC tasks should work out of the box, without the need to write an app_config.xml file. In fact, I never wrote one in Linux or Windows, only once in SETI@home I copied one and then went back to stock applications. I am running SETI@home and Einstein@home both on CPUs and GPUs, plus climateprediction.net. I was running Test4Theory@home since November 2010 and it never failed. Since the LHC@home consolidation, all LHC@home tasks fail on my 3 PCs, except SixTrack.
I am only doing a test run with Atlas@home on the laptop HP PC with SuSE Linux Leap 42.2, VirtualBox 5.1.26 and BOINC 7.4.22.
Tullio

Toby Broom
Volunteer moderator
Send message
Joined: 27 Sep 08
Posts: 359
Credit: 78,419,513
RAC: 112,883
Message 31987 - Posted: 17 Aug 2017, 17:06:02 UTC - in response to Message 31984.

I agree that it should work without app_config.xml and I think it does. If it doesn't then we can get the admins to adjust the configuration.

can you make a new thread so this stay on topic?

gyllic
Send message
Joined: 9 Dec 14
Posts: 71
Credit: 821,963
RAC: 5,072
Message 32351 - Posted: 8 Sep 2017, 14:40:48 UTC

these two have only 2 days deadline:
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=74431041
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=74435643

Crystal Pellet
Volunteer moderator
Volunteer tester
Send message
Joined: 14 Jan 10
Posts: 329
Credit: 2,779,814
RAC: 3,567
Message 32352 - Posted: 8 Sep 2017, 17:31:34 UTC - in response to Message 32351.

these two have only 2 days deadline:
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=74431041
https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=74435643

I suppose the original mentioned deadline change was meant for the ATLAS VM-tasks and not for the ATLAS native tasks, it seems.

Juha
Send message
Joined: 22 Mar 17
Posts: 19
Credit: 192,056
RAC: 914
Message 32353 - Posted: 8 Sep 2017, 17:39:08 UTC - in response to Message 32352.
Last modified: 8 Sep 2017, 17:40:55 UTC

I think admins have said that the VM and native tasks come from the same pool and therefore have the same deadlines.

edit: And in the first post it says one week. Guess they have really urgent tasks, then. :)

Message boards : News : Deadline change for ATLAS jobs