Message boards : ATLAS application : Avoid ATLAS task with app_config
Message board moderation

To post messages, you must log in.

AuthorMessage
Guiri-One[Andalucia]

Send message
Joined: 1 Feb 06
Posts: 66
Credit: 9,723
RAC: 0
Message 44292 - Posted: 11 Feb 2021, 9:33:26 UTC

Hi team,

Is there any way, like setting max CPU = 0, to avoid crunching ATLAS tasks?

My machine has good CPU but little ram and they come, I am lost :)

Thx in advanceĀ”

I tried with no luck, running only theory:

<app>
<name>Theory</name>
<app_version>
<app_name>Theory</app_name>
<plan_class>vbox64_mt_mcore_theory</plan_class>
<avg_ncpus>2.0</avg_ncpus>
<cmdline>--nthreads 2</cmdline>
</app_version>
</app>

<app>
<name>ATLAS</name>
<app_version>
<app_name>ATLAS</app_name>
<plan_class>vbox64_mt_mcore_atlas</plan_class>
<avg_ncpus>0</avg_ncpus>
<cmdline>--nthreads 2</cmdline>
</app_version>
</app>
ID: 44292 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 674
Credit: 43,151,503
RAC: 15,790
Message 44294 - Posted: 11 Feb 2021, 11:58:20 UTC

You can de-select Atlas in your project preferences here: https://lhcathome.cern.ch/lhcathome/prefs.php?subset=project. Just make the modification to the same venue (Home, School, Work etc.) that applies to your Host. Remember to also deselect 'If no work for selected applications is available, accept work from other applications?' as well.
ID: 44294 · Report as offensive     Reply Quote
Guiri-One[Andalucia]

Send message
Joined: 1 Feb 06
Posts: 66
Credit: 9,723
RAC: 0
Message 44296 - Posted: 11 Feb 2021, 12:07:32 UTC - in response to Message 44294.  

Hey,

Thx for the help. But I am behing a pool and I cant "login" and manipulate my preferences. I need tof orce it somehow locally, as I do with % of CPUS by overriding preferences.
ID: 44296 · 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: 222,929,849
RAC: 137,649
Message 44298 - Posted: 11 Feb 2021, 12:37:42 UTC - in response to Message 44296.  

... manipulate my preferences. I need tof orce it somehow locally ...

There's no way to configure this locally.
Not even if you would use a well formed app_config.xml following this page:
https://boinc.berkeley.edu/wiki/Client_configuration
ID: 44298 · Report as offensive     Reply Quote
marmot
Avatar

Send message
Joined: 5 Nov 15
Posts: 144
Credit: 6,301,268
RAC: 0
Message 44512 - Posted: 18 Mar 2021, 7:59:33 UTC
Last modified: 18 Mar 2021, 8:20:27 UTC

If you are creative there should be a way to prevent ATLAS from running locally.
I've stopped WU's locally 3 years back with some techniques on various projects but the details are fading from my brain.

The 1st technique was discovered accidentally by attempting to fit the most Theory WU's into limited RAM on my new server. I set the RAM size for the LHC WU's too small to succeed (there are other posts in the forums about how to use app_config.xml to adjust RAM used by Theory or ATLAS).
However, this leads to a long list of failed WU's but the server should eventually stop sending WU's (for the day, maybe longer I don't remember).

The more subtle method I had to invent was because LHC@Home would continue to d/l the ATLAS virtual disk even though I had refused ATLAS WU's in the preferences and had set BOINC to leave at least 3GB free.
LHC@Home ignored all my limits on it's behavior and was running my laptop hardrive out of space causing me data losses.

This procedure is for Windows OS:
1) Shut down BOINC
2) enter the LHC project data folder
3) rt click on the ATLAS_vbox_2.00_image.vdi file and select rename
4) copy the name into the copy-paste buffer
5) delete ATLAS_vbox_2.00_image.vdi
6) rt click on the folder space and choose NEW FILE
7) create a new text document
8) rt click that document and rename it to ATLAS_vbox_2.00_image.vdi by pasting the name from the copy-paste buffer (or use your brain's memory)
9) rt click on that new file and select Properties
10) set the file to Read Only.
11) You'll have a 0 byte invalid ATLAS image that can't be modified by BOINC

Restart BOINC.
Now ATLAS should fail to ever start because it fails to d/l it's image and the one marked Read Only fails checksums and obviously can't start a VM.
ID: 44512 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,733,230
RAC: 233,753
Message 44516 - Posted: 18 Mar 2021, 17:39:16 UTC

What about this:

<app>
<name>ATLAS</name>
<max_concurrent>-1</max_concurrent>
</app>
ID: 44516 · Report as offensive     Reply Quote
tullio

Send message
Joined: 19 Feb 08
Posts: 708
Credit: 4,336,250
RAC: 0
Message 44517 - Posted: 18 Mar 2021, 19:01:11 UTC

All my Atlas tasks perform beautifully. as my Theory tasks.Only CMS tasks fail every time, for reasons nobody can explain. I am not overclocking anything, all my setttings are standard.
Tullio
ID: 44517 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Nov 14
Posts: 602
Credit: 24,371,321
RAC: 0
Message 44518 - Posted: 18 Mar 2021, 20:46:21 UTC - in response to Message 44517.  

Only CMS tasks fail every time, for reasons nobody can explain.

Have you tried disabling your AV (not just adding BOINC Data to the exception list)?
ID: 44518 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,733,230
RAC: 233,753
Message 44521 - Posted: 19 Mar 2021, 22:04:00 UTC - in response to Message 44517.  

it strange, yes, looking it say it successfully ran the WU but then couldn't successfully power off the VM, so you successfully did work for CMS, just the project didn't think so.

Sometimes I get junk VM's left over that can cause problems, so I clear them out. How is your storage, could be it too slow clean up after the WU is done and then it time out boinc detecting. Last one I see is that if I run 1 WU/core then it can really slow my computer so I do use at most 80%, it still loads my CPU to 100% normally but it not red lined.
ID: 44521 · Report as offensive     Reply Quote
tullio

Send message
Joined: 19 Feb 08
Posts: 708
Credit: 4,336,250
RAC: 0
Message 44556 - Posted: 26 Mar 2021, 13:23:51 UTC - in response to Message 44518.  

Only CMS tasks fail every time, for reasons nobody can explain.

Have you tried disabling your AV (not just adding BOINC Data to the exception list)?

Yes I have disabled McAfee AV and CMS tasks fall the same. I am using only Windows Defender and I don't want to disable it.
Tullio
ID: 44556 · Report as offensive     Reply Quote
marmot
Avatar

Send message
Joined: 5 Nov 15
Posts: 144
Credit: 6,301,268
RAC: 0
Message 44568 - Posted: 26 Mar 2021, 23:08:06 UTC - in response to Message 44521.  
Last modified: 26 Mar 2021, 23:08:59 UTC

it strange, yes, looking it say it successfully ran the WU but then couldn't successfully power off the VM, so you successfully did work for CMS, just the project didn't think so.

Sometimes I get junk VM's left over that can cause problems, so I clear them out. How is your storage, could be it too slow clean up after the WU is done and then it time out boinc detecting. Last one I see is that if I run 1 WU/core then it can really slow my computer so I do use at most 80%, it still loads my CPU to 100% normally but it not red lined.


I've had to go into the media manager and delete hundreds of failed attached disks when something like this happened before.

IIRC ended up performing a complete uninstall of Oracle vBox and then a fresh install to get the problem to disappear.

(But this is off topic from the original poster's question about how to stop ATLAS from even running)
ID: 44568 · Report as offensive     Reply Quote

Message boards : ATLAS application : Avoid ATLAS task with app_config


©2024 CERN