61)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38447)
Posted 26 Mar 2019 by bronco Post: [quoteI avoid sixtrack. It is too easy, and requires no special software. Anyone can run it, so I let them.{/quote] Good point. It illustrates the lost opportunity cost concept very well... when you crunch sixtrack you lose the opportunity to crunch a task that many other volunteers cannot crunch. |
62)
Message boards :
CMS Application :
New Version v49.00
(Message 38444)
Posted 26 Mar 2019 by bronco Post: This new version updates the CVMFS cache to reduce the downloads each time the VM starts. Nice! Native CMS by August? |
63)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38443)
Posted 26 Mar 2019 by bronco Post: Hoping to add native CMS to the mix someday. Optimism promotes longevity and it tastes better than liver, kale and fat free ice cream. If you haven't already done so, snag some sixtrack and see how nicely they play with native ATLAS and native Theory. With "switch between tasks every..." set to 2080 minutes and a sane task cache nothing gets preempted which avoids ATLAS restarting from 0 events. Very nice!! |
64)
Message boards :
Theory Application :
(Native) Theory - Sherpa looooooong runners
(Message 38442)
Posted 26 Mar 2019 by bronco Post: Suddenly it was over very quickly: https://lhcathome.cern.ch/lhcathome/result.php?resultid=219941269 Wonderful! Your point is? |
65)
Message boards :
Theory Application :
(Native) Theory - Sherpa looooooong runners
(Message 38440)
Posted 26 Mar 2019 by bronco Post: Still running: According to the Sherpa 2.1.0 manual, "Sherpa will then move on to integrate the other processes specified in the run card." And "When the integration is complete, the event generation will start." |
66)
Message boards :
Theory Application :
(Native) Theory - Sherpa looooooong runners
(Message 38437)
Posted 26 Mar 2019 by bronco Post: All you can do is abort the task. Really? If the task fails (and his task most certainly will), it won't even upload a result. No result = no science. If he had aborted the task 24 hours ago he could have received a pythia, herwig, <whatever> that is far more likely to succeed and do some worthwhile science. It has been stated that they learn something even if the job fails. Really? What do they learn? All they learn is that the job failed. They can learn that fact from the 2 failures from the 2 wingmen. It's the principle of lost opportunity cost... every failed job is a lost opportunity to do some useful science. Have a Sherpa running for 10K Minutes now And show every minute a answer in runRivet.log.You got lucky. In Theory-Thread is a link for the Sherpa Documentation.Please post that link, I couldn't find it but I would like to read it. |
67)
Message boards :
Theory Application :
(Native) Theory - Sherpa looooooong runners
(Message 38434)
Posted 26 Mar 2019 by bronco Post: Display update finished (0 histograms, 0 events). runRivet.log is the right file and the task itself is working. The sherpa job is most certainly a fail. The giveaway is that it's stuck at 0 histograms and 0 events. All you can do is abort the task. |
68)
Message boards :
Theory Application :
Installation of CVMFS
(Message 38411)
Posted 24 Mar 2019 by bronco Post: The server scheduler is sometimes confused, how to interpret the max # settings. You need root privilege to create/edit that file otherwise it's read only I usually do sudo nano ../app_config.xml. Nano is a minimal editor but it works well enough for short files. |
69)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38403)
Posted 24 Mar 2019 by bronco Post: I am now getting both Native Theory and Native ATLAS with the appropriated preferences selected. Yes, very well. I have a native ATLAS and a native Theory running concurrently on a host with only 4 GB RAM :) Hoping to add native CMS to the mix someday. |
70)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38386)
Posted 23 Mar 2019 by bronco Post: No result so far. Sherpa 1.4.3 for now 25 hours running and hoping to get a good end therefore. Now I see I was wrong :-( Die, sherpa, die!! Setting my watchdog script to immediately abort any sherpa less than version 5.0.0. |
71)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38365)
Posted 21 Mar 2019 by bronco Post: Maybe you are the first one with a success ;) @maeax I think CP is sherpa shaming you ;) Don't give in, keep the faith Go, sherpa, go!! |
72)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38356)
Posted 21 Mar 2019 by bronco Post: No result so far. Sherpa 1.4.3 for now 25 hours running and hoping to get a good end therefore. Yes, I understand. Sherpa has earned a bad reputation perhaps unfairly. Maybe it just needs more time than it is allowed in the VBox tasks. I will repeat your test next time I get a sherpa. Thank you for pointing the right direction :) Go, sherpa, go!! |
73)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38351)
Posted 20 Mar 2019 by bronco Post: No result so far. Sherpa 1.4.3 for now 25 hours running and hoping to get a good end therefore. Death by graceful shutdown must end! Sherpa lives matter too!! |
74)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38339)
Posted 20 Mar 2019 by bronco Post: The task is set to use 2 CPUs by default and barely over 1 is used and the reported time has run time = exactly CPU time. To the second on every task. At most I see 1.5 cores when the task is really short. 6min run time, 8 min CPU time.Don't trust the values reported in the results, specially when they are equal. I had the same problem. I traced the cause to an error in my app_config.xml. Maybe you made the same error I did? To avoid adding "noise" to this thread I posted a proper block for native Theory in a separate thread at https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4975 |
75)
Message boards :
Theory Application :
app_config.xml block for native Theory
(Message 38338)
Posted 20 Mar 2019 by bronco Post: Should look similar to the following, <app> <name>TheoryN</name> <max_concurrent>4</max_concurrent> </app> <app_version> <app_name>TheoryN</app_name> <plan_class>native_theory</plan_class> <avg_ncpus>1</avg_ncpus> <cmdline>--nthreads 1 </cmdline> <!-- RAM formula: RAM = ? --> <cmdline>--memory_size_mb 400 </cmdline> </app_version> |
76)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38320)
Posted 19 Mar 2019 by bronco Post: Tried "sudo touch /var/lib/boinc/slots/0/cernvm/shared/shutdown" to gracefully kill a hung sherpa job. The command succeeded but the task didn't shutdown. Is graceful shutdown not an option with native Theory or am I creating the file in the wrong folder? |
77)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38297)
Posted 19 Mar 2019 by bronco Post: ... I don't like the CPU's being used only partially but I'll ignore it if they promise there won't be any sherpa jobs.Promise: sherpa's will come. I've one running at the moment ;) Damn! Guess I have to modify my watchdog script again. |
78)
Message boards :
Number crunching :
Checklist Version 3 for Atlas@Home (and other VM-based Projects) on your PC
(Message 38295)
Posted 19 Mar 2019 by bronco Post: Does it make sense to try all that on a box with 8Gig of RAM? I used to do it on 8GB. At first I got a lot of invalids. Then I boosted the "switch between tasks every..." setting to 24 hours and got a considerably higher success rate (but not 100%). I also learned to schedule OS updates and reboots around the ATLAS tasks so as not to suspend them. In theory ATLAS VBox tasks should not be bothered by suspending/resuming but in practice I found they are. YMMV. |
79)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38288)
Posted 19 Mar 2019 by bronco Post: The task is set to use 2 CPUs by default and barely over 1 is used and the reported time has run time = exactly CPU time. To the second on every task. At most I see 1.5 cores when the task is really short. 6min run time, 8 min CPU time.Don't trust the values reported in the results, specially when they are equal. OK it makes sense now with respect to the numbers adding up correctly. I don't like the CPU's being used only partially but I'll ignore it if they promise there won't be any sherpa jobs. |
80)
Message boards :
Theory Application :
Issues Native Theory application
(Message 38285)
Posted 19 Mar 2019 by bronco Post: The setup went without error, thanks Ivan for the great directions.The directions are actually by Laurence ;) But BOINC manager shows 2 X 2-CPU tasks = 4 CPU's in use, in other words no idle CPU's. Also, the task run times are nearly equal to the task CPU times when I would expect CPU time to be a little less than double the run time. |
©2024 CERN