Message boards :
ATLAS application :
Tasks (2-core) failing after 12-13 minutes
Message board moderation
Author | Message |
---|---|
Send message Joined: 18 Dec 15 Posts: 1688 Credit: 103,753,631 RAC: 122,148 |
This afternoon, several 2-core ATLAS Tasks were failing after about 12-13 minutes. As an example, please see here: https://lhcathome.cern.ch/lhcathome/result.php?resultid=127234375 can anyone tell me what's going wrong? FYI, on this machine, during the past year I have been cruching numerous 3-core ATLAS tasks without any problems, also a few new ones under the LHC@home roof. So, why would 3-core tasks work well, and 2-core tasks don't ? Whatever this Information is worth: also this afternoon, several CMS tasks (1-core) failed after about the same time. |
Send message Joined: 9 Dec 14 Posts: 202 Credit: 2,533,875 RAC: 0 |
your log shows: 2017-03-20 17:53:45 (6816): Setting Memory Size for VM. (3600MB) according to this post https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4164&postid=29445 you have to increase the memory for the VM to at least 4400MB for a multicore task (by app_config for example) |
Send message Joined: 18 Dec 15 Posts: 1688 Credit: 103,753,631 RAC: 122,148 |
... you have to increase the memory for the VM to at least 4400MB for a multicore (by app_config for example) thanks for the quick Information. What I am surprised, though, is that on this 32 GB RAM machine I never before had to manually increase the VM Memory. Would the VM not take as large a memory as needed for a given task? (as long as there is enough RAM left)? Even 3-core ATLAS tasks (from the former ATLAS@home-server as well as recently from the LHC@home Server)did NOT require any manual adaption of the VM memory size. So, why would this now suddenly be necessary with 2-core tasks? |
Send message Joined: 18 Dec 15 Posts: 1688 Credit: 103,753,631 RAC: 122,148 |
Out of interest, on older stderr files I now checked the VM memory creations for 1-core thru for 3-core ATLAS Tasks: for 1-core: 2600 MB for 2-core: 3600 MB for 3-core: 4600 MB As said above, I would guess that VM is knowing what it's doing. Or is there a bug somewhere in the VM and the 3600 MB for 2-core is indeed not enough? With my total 32GB RAM, I would have not problem at all to manually (by app_config, if I can find this somewhere) set the VM RAM to any higher value than 3600 MB. |
Send message Joined: 9 Dec 14 Posts: 202 Credit: 2,533,875 RAC: 0 |
look here: https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4164&postid=29450 for your app config place a file called app_config.xml in the project directory and write this into it to run 6 atlas tasks with 2 cores each at the same time (if you have a 12 core processor) and 4400MB memory for each task (make sure to save it in the right format): <app_config> <app> <name>ATLAS</name> <max_concurrent>6</max_concurrent> </app> <app_version> <app_name>ATLAS</app_name> <avg_ncpus>2.000000</avg_ncpus> <plan_class>vbox64_mt_mcore_atlas</plan_class> <cmdline>--memory_size_mb 4400</cmdline> </app_version> </app_config> restart boinc and you should be good to go. |
Send message Joined: 18 Dec 15 Posts: 1688 Credit: 103,753,631 RAC: 122,148 |
@gyllic, many thanks for your help! If I am through with the one 3-core task and the other single-core tasks which are still running, I will make the Change and see what happens :-) |
Send message Joined: 18 Dec 16 Posts: 123 Credit: 37,495,365 RAC: 0 |
On ATLAS@Home the default rule for setting the VirtualBox memory was: 2.5 GB + 0.8 * ncores (see Yeti's checklist). On LHC@Home, after the introduction of a new simulator version 1.01 (more efficient), the rule was changed to 1.6 GB + 1 * ncores (see https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4146&postid=29260#29260). This explains the figures you found in your stderr files, Erich. Tests with the new version showed that the rule was OK for 3 cores and above, but not OK for 2 cores. So if you want to run 2-cores, you need to configure an app_config.xml file with 4400 MB to overwrite the default 3600 MB, as mentionned by gyllic. We are the product of random evolution. |
©2024 CERN