Message boards : ATLAS application : ATLAS tasks couldn't run
Message board moderation

To post messages, you must log in.

AuthorMessage
16492

Send message
Joined: 13 Jul 19
Posts: 2
Credit: 70,366
RAC: 1
Message 39395 - Posted: 19 Jul 2019, 23:15:41 UTC

When runing a Atlas task, BONIC always displays waiting for memory

    Application
    ATLAS Simulation 1.01 (vbox64_mt_mcore_atlas)
    Name
    tQMKDm7Jo8unsSi4apGgGQJmABFKDmABFKDmaNWQDmABFKDmlUV4Oo
    State
    Waiting for memory
    Received
    20/07/2019 06:31:25
    Report deadline
    27/07/2019 06:31:24
    Resources
    4 CPUs
    Estimated computation size
    43,200 GFLOPs
    CPU time
    ---
    CPU time since checkpoint
    ---
    Elapsed time
    ---
    Estimated time remaining
    07:01:21
    Fraction done
    0.000%
    Virtual memory size
    6.20 MB
    Working set size
    6.45 GB
    Directory
    slots/0
    Executable
    vboxwrapper_26196_windows_x86_64.exe

ID: 39395 · Report as offensive     Reply Quote
bronco

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,252,104
RAC: 9,833
Message 39396 - Posted: 19 Jul 2019, 23:34:32 UTC - in response to Message 39395.  

If you unhide your computers it will be much easier to look through your result logs and diagnose the problem.
ID: 39396 · Report as offensive     Reply Quote
16492

Send message
Joined: 13 Jul 19
Posts: 2
Credit: 70,366
RAC: 1
Message 39397 - Posted: 19 Jul 2019, 23:57:54 UTC - in response to Message 39396.  

Now I unhided my computer.
ID: 39397 · Report as offensive     Reply Quote
computezrmle
Avatar

Send message
Joined: 15 Jun 08
Posts: 1141
Credit: 56,182,403
RAC: 96,674
Message 39400 - Posted: 20 Jul 2019, 8:14:32 UTC

Running ATLAS on a host whith lots of cores (here: 12) but only 8 GB RAM might not be a good idea to start with.
The VM started by each task will allocate 6600 MB RAM by default (4-core setup) and the log shows that the BOINC client paused and restarted the task several times:
https://lhcathome.cern.ch/lhcathome/result.php?resultid=237234947
Those restarts are extremely inefficient and very often they end with unsuccessful tasks.

It may be better to deselect ATLAS and focus on SixTrack and Theory on this host.
Theory may be limited to a 2-core setup using the following app_config.xml:
<app_config>
  <app_version>
    <app_name>Theory</app_name>
    <plan_class>vbox64_mt_mcore</plan_class>
    <avg_ncpus>2.0</avg_ncpus>
    <cmdline>--memory_size_mb 830</cmdline>
  </app_version>
</app_config>

"Read config files" from the menue must be run or the BOINC client has to be restarted or to activate the changes.
Already running tasks will not use the changes.

The format of an app_config.xml is specified here:
https://boinc.berkeley.edu/wiki/Client_configuration
Other parameters than specified will be ignored by the BOINC client, hence it makes no sense to specify <max_ncpus> like in GreenHat's post.
The nthreads setting there is also redundant as the BOINC client will correctly configure the #cores for the VM according to <avg_ncpus>.


@GreenHat
Especially for inexperienced volunteers it's hard to follow your advice without having a chance to check them against your logfiles.
In addition mcplots does only show computers that are attached to Theory.
Computers that only run ATLAS or SixTrack are not shown there.
Hence it is highly recommended that you make your hosts visible.

It would also be nice if you could mark an URL as URL.
Just highlight it and press the editor's URL button:
GreenHat wrote:
More info: https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4178#29560
ID: 39400 · Report as offensive     Reply Quote

Message boards : ATLAS application : ATLAS tasks couldn't run


©2019 CERN