Multiple VM workunits running at the same time?

Everything about the project RNA World
Nachricht
Autor
IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Multiple VM workunits running at the same time?

#1 Ungelesener Beitrag von IanEdwardJames » 29.07.2015 10:41

Hi,

Is there a way to get 2 or more VM workunits running at the same time or is it system performance related/restricted?

I am currently running a monster (estimated 1700 hours remaining (RNA estimation is 16 week run time))

But I have another sitting there waiting to run but it won't start. It is also a long one (estimated 1200 hours)

So wondering if there is a configuration issue, a system performance related issue or if Vbox can only run one thing anyway so I just have to wait.

It would be nice to make a start though, if its going to take so long.



Operating System: Windows 7 Professional 64-bit (6.1, Build 7601) Service Pack 1 (7601.win7sp1_gdr.150525-0603)
Processor: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz (8 CPUs), ~3.1GHz
Memory: 6144MB RAM
Available OS Memory: 6142MB RAM
Page File: 10098MB used, 2184MB available

System usually sits around 4.3 GB Ram used.

Thanks guys,

Ian

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Multiple VM workunits running at the same time?

#2 Ungelesener Beitrag von ChristianB » 29.07.2015 11:06

Hi Ian,

every RNA VM task needs to have 3 GB of free RAM available at all times even if it doesn't use all of it. But it's possible that it will use all of the 3 GB that BOINC reserved for it. So in order to run two VM tasks at the same time you need to have at least 6 GB of RAM available to BOINC that is not used by the OS. In your case it's imposible to run 2 VMs with only 6 GB RAM total.

Because BOINC doesn't know about the deadline change it thinks it can still run the second task because the first one has to finish soon. That's wrong but we (or you) have no way of letting the BOINC client know it's error.

IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Re: Multiple VM workunits running at the same time?

#3 Ungelesener Beitrag von IanEdwardJames » 30.07.2015 10:37

Thanks for your reply Christian, you have answered my question well.

I might look at getting some more RAM maybe.

Ian

IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Re: Multiple VM workunits running at the same time?

#4 Ungelesener Beitrag von IanEdwardJames » 05.02.2016 23:06

Ok, I now have 12 Gb of RAM and system is running 3 VM tasks. The 3rd of which it claims it is running but the percentage completed is not increasing past 0.1% and the RAM usage (around 7Gb) does not really indicate that 3 VM tasks are running. Boinc has access to 95% (11.4 GB) of total RAM at all times. I would anticipate that it will require around 10GB, which leaves approx 1.4 Gb of headroom.
Again just checking that there is no other configuration issues that need to be changed to allow 3 VM taks to run in Boinc or VBOX?

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Multiple VM workunits running at the same time?

#5 Ungelesener Beitrag von Jacob Klein » 06.02.2016 05:12

The RAM usage is not a good indicator of a task running - each RNA World VM task uses "up to 3 GB", and BOINC must "consider 3 GB reserved per task" (per VirtualBox rules), but the task itself may be using much less, like only 1 GB.

Instead, it's better to monitor Task Manager's process tab, "CPU" column... and then also monitor the "progress.txt" files within the slots/shared folders of the tasks (which should be updating every minute or so, unless at 0.98765, in which case it is normal to see progress.txt stay at that value indefinitely).

I notice from your computer list:
http://www.rnaworld.de/rnaworld/results ... stid=19092

- 2 of the tasks are: cmsearch VM (VirtualBox) 1.0.2 v1.15 (vbox64)
- 1 of the tasks is: cmsearch VM (VirtualBox) 1.0.2 v1.16 (vbox64)

Is the v1.16 one the task that's not behaving as you expect? Does Task Manager and progress.txt indicate it is progressing?

IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Re: Multiple VM workunits running at the same time?

#6 Ungelesener Beitrag von IanEdwardJames » 06.02.2016 07:51

Yes it is the 1.0.2 v1.16 that I am looking at.

Progress is increasing in the text file and there are 3x Vboxheadless.exe all running around the same cpu cycles so I will take this as it is running ok. I could not differentiate between the 2 different types in task manager but thought seeing as there are 3 and they are all similar it would be good.

:)

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Multiple VM workunits running at the same time?

#7 Ungelesener Beitrag von Jacob Klein » 06.02.2016 07:55

Well, there may be a "progress display" bug, but it sounds like the task itself is running correctly. I've been trying for weeks to get a new task (to help test v1.16) -- consider yourself lucky to have it!

IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Re: Multiple VM workunits running at the same time?

#8 Ungelesener Beitrag von IanEdwardJames » 06.02.2016 13:32

That's all good then, didn't mean to be complaining sorry, just wanted to make sure that it was running if it was able to.

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Multiple VM workunits running at the same time?

#9 Ungelesener Beitrag von ChristianB » 06.02.2016 14:19

This is a bit alarming. If there really is a new bug in vboxwrapper when I just upgraded to the latest and greatest and there are so much iterations of the vboxwrapper in between. I have some logfiles for some other snapshot related problem that I wanted to send to Rom, I guess I have to include this now too.

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Multiple VM workunits running at the same time?

#10 Ungelesener Beitrag von Jacob Klein » 06.02.2016 15:14

For reference, when I turn on the "app_msg_receive" debug flag in cc_config.xml, I see the following for my v1.15 tasks:

Code: Alles auswählen

2/6/2016 9:11:50 AM | RNA World | [app_msg_receive] got msg from slot 18: <current_cpu_time>1.063219e+007</current_cpu_time><checkpoint_cpu_time>1.063105e+007</checkpoint_cpu_time><fraction_done>0.000000e+000</fraction_done>
2/6/2016 9:11:50 AM | RNA World | [app_msg_receive] got msg from slot 17: <current_cpu_time>2.081030e+006</current_cpu_time><checkpoint_cpu_time>2.080301e+006</checkpoint_cpu_time><fraction_done>0.000000e+000</fraction_done>
2/6/2016 9:11:50 AM | RNA World | [app_msg_receive] got msg from slot 25: <current_cpu_time>4.710523e+006</current_cpu_time><checkpoint_cpu_time>4.709913e+006</checkpoint_cpu_time><fraction_done>0.000000e+000</fraction_done>
... fraction_done is 0 for the RNA World v1.15 tasks
... and the Progress % shown in BOINC does not at all match the progress % found in the Progress.txt file.

So, it sounds like v1.16 has its own kind of new progress display problem, but v1.15 didn't work right either :)

IanEdwardJames
Mikrocruncher
Mikrocruncher
Beiträge: 23
Registriert: 15.03.2011 10:40

Re: Multiple VM workunits running at the same time?

#11 Ungelesener Beitrag von IanEdwardJames » 07.02.2016 09:52

Well let me know if I can do anything to help, I am manually keeping track of hours run/ hours remaining / % complete and now started recording the progress.txt result for that WU. I'll just monitor it for a few days but if I can do anything else to help, just sing out.

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Multiple VM workunits running at the same time?

#12 Ungelesener Beitrag von Jacob Klein » 07.02.2016 19:02

Ian:

Can you please do the following:
For each of your 3 tasks, after letting them run for at least 5 minutes, please list:
- the Application version number
- the executable/process in the Task Properties
- the Elapsed time
- the CPU time at last checkpoint in the Task Properties
- the CPU time
- the Time Remaining (Estimated)

Does it seem that the v1.16 task shows a bad/malformed "Time Remaining (Estimated)" value? Note: Even if it does, the task is still progressing normally, and should NOT be aborted!

I've been doing some local testing with the various VBoxWrapper versions, and I reproduced your "constant 0.1% Progress" bug, but for me, there was an additional side effect where the "Time Remaining (Estimated)" was completely incorrect/hosed/malformed, and I'm wondering if you have the same issue.

Thanks!
Jacob Klein

Antworten

Zurück zu „RNA World Discussions (english)“