Message boards : ATLAS application : WU uses wrong number of threads
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 436
Credit: 117,893,361
RAC: 3,475
Message 37933 - Posted: 5 Feb 2019, 16:52:46 UTC

My client is configured to run 4-Core-WUs, but this WU uses 8 Threads (Cores):




Supporting BOINC, a great concept !
ID: 37933 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 2027
Credit: 148,464,649
RAC: 118,224
Message 37934 - Posted: 5 Feb 2019, 18:31:34 UTC

A very old error type.
It is caused by a misconfigured task (or complete batch) and can only be fixed by the project team.

If the VM has enough RAM to satisfy all running threads the tasks will most likely finish successfully but you may notice an efficiency drop.
ID: 37934 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 590
Credit: 33,926,216
RAC: 18,819
Message 37940 - Posted: 6 Feb 2019, 7:44:56 UTC

Am I reading it wrong or does the ~50 % CPU usage indicate that it is actually using 4 cores and running 2 athena.py processes on each? But it is a fault anyway.
ID: 37940 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 2027
Credit: 148,464,649
RAC: 118,224
Message 37941 - Posted: 6 Feb 2019, 8:25:29 UTC - in response to Message 37940.  

... does the ~50 % CPU usage indicate that it is actually using 4 cores and running 2 athena.py processes on each?

That's exactly what it shows.

As Yeti wrote it was a 4-core setup running 8 threads (and meanwhile it finished successfully):
2019-02-05 17:07:29 (24100): Setting CPU Count for VM. (4)
.
.
.
2019-02-05 21:42:12 (24100): VM Completion File Detected.
2019-02-05 21:42:12 (24100): Powering off VM.
2019-02-05 21:42:13 (24100): Successfully stopped VM.

From the host's perspective there shouldn't have been any negative observations.
The OS inside the VM had to handle lots of context switches hence a performance penalty.
ID: 37941 · Report as offensive     Reply Quote
Jonathan

Send message
Joined: 25 Sep 17
Posts: 76
Credit: 2,089,097
RAC: 220
Message 37973 - Posted: 11 Feb 2019, 4:13:53 UTC - in response to Message 37941.  

I have a currently running Atlas task doing the same thing. Yeti's log for that work unit is a bit strange, see below. It looks like it was trying to set ATHENA_PROC_NUMBER=4 but ???

2019-02-05 17:10:31 (24100): Guest Log: Copying inCpuotp yfiinlge si nipnutto fRiulneAst lianst.o
2019-02-05 17:10:31 (24100): Guest Log: tlas.
2019-02-05 17:10:31 (24100): Guest Log: Copied input files into RunAtlas.
2019-02-05 17:10:31 (24100): Guest Log: Copied input files into RunAtlas.
2019-02-05 17:14:54 (24100): Guest Log: cocpieodp itheed wtehbea pwpe btaopp /tvoa r//vwwawr/
2019-02-05 17:14:54 (24100): Guest Log: w
2019-02-05 17:14:54 (24100): Guest Log: TThhiiss vvmm ddooeess nnoott nneeeedd ttoo sseettuupp hhttttpp pprrooxxyy
2019-02-05 17:14:54 (24100): Guest Log: TThhiiss vvmm ddooeess nnoott nneeeedd ttoo sseettuupp hhttttpp pprrooxxyy
2019-02-05 17:14:54 (24100): Guest Log: ATTHHEENNAA__PPRROOCC__NNUUMMBBEERR==44
2019-02-05 17:14:54 (24100): Guest Log: ATTHHEENNAA__PPRROOCC__NNUUMMBBEERR==44
2019-02-05 17:14:54 (24100): Guest Log: StartinSg tATrLtAiSn gj oAbT.LAS (jPoanbd.a I(PDa=n4d2a3I4D9=2442439459 2t4a4s9kI5D t=a1s6k9I3D4=18655)34
2019-02-05 17:14:54 (24100): Guest Log: )
ID: 37973 · Report as offensive     Reply Quote

Message boards : ATLAS application : WU uses wrong number of threads


©2022 CERN