Message boards :
ATLAS application :
ATLAS vbox and native 3.01
Message board moderation
Previous · 1 · 2 · 3 · 4 · Next
Author | Message |
---|---|
Send message Joined: 15 Jun 08 Posts: 2531 Credit: 253,722,201 RAC: 41,981 |
... ALT-F2 monitoring ... Right, it works here since the tasks are still processing data from LHC Run 2 using the matching scripts. It didn't work at -dev recently because the tasks there were tests from LHC Run 3 (data/scripts). ATLAS Monitoring is not yet prepared to deal with the slightly modified Run 3 logging. Once this is implemented it will be tested at -dev first. So far the suggestion is to stay patient. |
Send message Joined: 25 Sep 17 Posts: 99 Credit: 3,425,566 RAC: 0 |
Is the VM RAM size going to be adjusted eventually or will the current defaults be used because 3.01 can run 'Run 2' and 'Run 3'? Will the VM change the RAM assigned by 'Run 2' vs 'Run 3'? |
Send message Joined: 14 Jan 10 Posts: 1417 Credit: 9,441,051 RAC: 798 |
Is the VM RAM size going to be adjusted eventually or will the current defaults be used because 3.01 can run 'Run 2' and 'Run 3'? Will the VM change the RAM assigned by 'Run 2' vs 'Run 3'? The RAM size is set to 4000 MB independend of running tasks from run 2 or run 3. It also doesn't matter whether you run single core vbox tasks or multi-core e.g. 4 or 8 cores. |
Send message Joined: 25 Sep 17 Posts: 99 Credit: 3,425,566 RAC: 0 |
VM is assigning 8400 MB for my 6 core tasks. Max # CPUs set to 6 on LHC preferences. No app_config.xml used. |
Send message Joined: 17 Sep 04 Posts: 105 Credit: 32,824,853 RAC: 579 |
Is the VM RAM size going to be adjusted eventually or will the current defaults be used because 3.01 can run 'Run 2' and 'Run 3'? Will the VM change the RAM assigned by 'Run 2' vs 'Run 3'? Was it determined in years past that runs using 4 or more cores resulted in a sharp reduction in processing efficiency? Regards, Bob P. |
Send message Joined: 28 Sep 04 Posts: 728 Credit: 48,901,114 RAC: 21,856 |
VM is assigning 8400 MB for my 6 core tasks. Max # CPUs set to 6 on LHC preferences. No app_config.xml used. I am running Atlas tasks with VM in Windows 10 using 4 cores per task. I removed the commandline parameter from my app_config.xml that set the memory usage to 6600 KB. The tasks still use 6600 KB memory with 4 CPUs when using 3.01 application. This can be seen also in Alt+F3 TOP terminal window. |
Send message Joined: 28 Sep 04 Posts: 728 Credit: 48,901,114 RAC: 21,856 |
VM is assigning 8400 MB for my 6 core tasks. Max # CPUs set to 6 on LHC preferences. No app_config.xml used. Edit: I meant of course 6600 MB |
Send message Joined: 28 Sep 04 Posts: 728 Credit: 48,901,114 RAC: 21,856 |
Is there a way to recognize the run 3 tasks? From the name? Memory consumption? Or how? |
Send message Joined: 14 Jan 10 Posts: 1417 Credit: 9,441,051 RAC: 798 |
Is there a way to recognize the run 3 tasks? From the name? Memory consumption? Or how? At the moment the output of ALT-F2 monitoring is garbled for run 3 and OK for run 2. |
Send message Joined: 16 Jun 20 Posts: 8 Credit: 2,318,092 RAC: 0 |
I'm running ATLAS on a Mac Mini (MacOS 12.6.3) under BOINC (7.20.4) and VirtualBox (7.0.6) with five CPUs, a max memory allocation of 12GB, and disc of 40GB. Each ATLAST task is set for 5 CPUs, and an initial run time of 00:22:20. They reach 100% after the prescribed time, then run away. I've just chopped the last batch as one had run from 00:22:20 to more than 19 hours while showing progress as 100%. All was fine until two days ago, when I upgraded VB from 7.0.4. However, my Ubuntu VMs run without problem Is there an LHC problem with VB 7.0.6? |
Send message Joined: 2 May 07 Posts: 2242 Credit: 173,902,375 RAC: 2,798 |
For Windows we need Python. |
Send message Joined: 15 Jun 08 Posts: 2531 Credit: 253,722,201 RAC: 41,981 |
... with five CPUs, a max memory allocation of 12GB, and disc of 40GB Where did you set this? The recent logs show 7500 MB are allocated for the ATLAS VM: https://lhcathome.cern.ch/lhcathome/result.php?resultid=391062360 Same here: task is set for ... an initial run time of 00:22:20 Where did you set this? ... had run from 00:22:20 to more than 19 hours Very unusual. ... while showing progress as 100% This might be caused by the fact that ATLAS published a new version. App updates always mess BOINC's runtime estimation. It will take a while until BOINC (server/client) negotiate the new effective GFLOPs value to be used in the runtime/credit calculation. Until then BOINC's progress bar and runtime estimation are pretty much useless. Beside that ATLAS 3.x has a new (larger) vdi file. Unfortunately your computer did not yet report a 3.x task, hence it's not possible to check the log for errors. LHC shouldn't have a problem with Vbox 7.x, especially since you successfully ran the previous version. |
Send message Joined: 25 Sep 17 Posts: 99 Credit: 3,425,566 RAC: 0 |
I don't see any aborted ATLAS tasks listed under your computer. Two CMS apps show aborted Did you set LHC to No New Tasks, abort non running tasks or allow the queue to clear before your VirtualBox upgrade? That is usually the best procedure. |
Send message Joined: 16 Jun 20 Posts: 8 Credit: 2,318,092 RAC: 0 |
Where did you set this? I didn't set it. That's the estimated process time when it arrives. My settings give BOINC a 40% share of 32GB while running native under MacOS. Very unusual. Maybe, but the next task is doing the same. True the percentage is not exactly 100, but it's 99.992, and I would expect that tomorrow morning, with an additional 12 hours on the clock, it might have increased a couple of thousandths of a percent. https://drive.google.com/file/d/1qDWN1Kdv82dz4bCAwN4EZ1SES4UWmzWq/view?usp=share_link |
Send message Joined: 16 Jun 20 Posts: 8 Credit: 2,318,092 RAC: 0 |
Where did you set this? my BOINC settings give it 40% of 32GB Task time is the estimate provided when the task arrives. Very unusual. Maybe so, but the next one is the same. Granted, it's 99.992% at the moment, but it's run for 3hrs 25mins out of its allotted 20 mins and 20 seconds, and looks as though it might accrue another couple of thousandths of a percent over the next twelve hours. https://drive.google.com/file/d/1qDWN1Kdv82dz4bCAwN4EZ1SES4UWmzWq/view?usp=share_link |
Send message Joined: 15 Jun 08 Posts: 2531 Credit: 253,722,201 RAC: 41,981 |
OK, I was thinking in a different direction. As already mentioned, don't trust the progress bar nor the remaining time until your computer returned a couple of results. Just let ATLAS run, ideally without intermediate suspension. |
Send message Joined: 16 Jun 20 Posts: 8 Credit: 2,318,092 RAC: 0 |
Here's the same task, 16 hours later. It has been showing remaining time as --- and percent complete as 100.000% for more than ten hours. The elapsed time is clocking up, but there is no CPU activity associated with the task. https://drive.google.com/file/d/199A_stcrLa-UzNMEwYPzLSPFfkhg8NFY/view?usp=share_link As fas as I'm concerned, it's a runaway. I can't afford to keep 45% of my available CPUs clogged with a series of runaway tasks, so I've chopped it. I'll come back to LHC in a few weeks to see if the problems are sorted. |
Send message Joined: 25 Sep 17 Posts: 99 Credit: 3,425,566 RAC: 0 |
I don't think the ATLAS task like to be started and stopped. I would pick a single project application to run and stick to only that for now. You can also work through Yeti's checklist. It's a sticky in the forums. Your picture shows an ATLAS task waiting to run at about 8 minutes in. It takes about 15 minutes before the computation phase of ATLAS as it is talking CVMFS and getting needed files and info. |
Send message Joined: 4 Mar 17 Posts: 25 Credit: 10,262,043 RAC: 7,610 |
Seems like the tasks since a few hours are some tasks 240MB again instead of the over 1GB. tasks with EVNT 321... are the smaller ones. there are still coming some 327... with 1GB. |
Send message Joined: 2 May 07 Posts: 2242 Credit: 173,902,375 RAC: 2,798 |
BellyNitpicker, your OS is Darwin and you have successful Tasks with the old Atlas-Version 2.03. With Version 3.01 something went wrong. maybe more RAM needed. |
©2024 CERN