Advanced search

Forums : Technical Support : camb_boinc2docker WUs failing
Message board moderation

To post messages, you must log in.

AuthorMessage
Kaddaman

Send message
Joined: 13 Oct 21
Posts: 7
Credit: 215,774
RAC: 1,120
Message 22853 - Posted: 14 Oct 2021, 19:34:44 UTC

Hey guys!
I just recently started crunching for Cosmology@Home and noticed that the multi-core WUs (camb_boinc2docker) always fail. This is what the log says:

14/10/2021 00:58:02 | Cosmology@Home | Starting task camb_boinc2docker_1671726_1634164736.239875_0
14/10/2021 00:58:25 | Cosmology@Home | Computation for task camb_boinc2docker_1671726_1634164736.239875_0 finished
14/10/2021 00:58:25 | Cosmology@Home | Output file camb_boinc2docker_1671726_1634164736.239875_0_r1680096054_0.tgz for task camb_boinc2docker_1671726_1634164736.239875_0 absent
Looking at the computer specs from Cosmology@Home account settings says "Virtualbox (6.1.26) installed, CPU has hardware virtualization support and it is enabled"

Anybody got any ideas? Is there a difference in RAC between docker and legacy? If docker generates more credit, I'd like to find a solution :P
ID: 22853 · Report as offensive     Reply Quote
maeax

Send message
Joined: 21 Dec 17
Posts: 30
Credit: 4,312,100
RAC: 26,033
Message 22855 - Posted: 14 Oct 2021, 23:58:25 UTC

VBoxManage.exe: error: Not in a hypervisor partition (HVP=0) (VERR_NEM_NOT_AVAILABLE).
VBoxManage.exe: error: AMD-V is disabled in the BIOS (or by the host OS) (VERR_SVM_DISABLED)
VBoxManage.exe: error: Details: code E_FAIL (0x80004005), component ConsoleWrap, interface IConsole
You may check your BIOS again, AMD-V is disabled.
Your CPU's are set to 12, maybe set it to One for the first time. (RAM is not enough?)
ID: 22855 · Report as offensive     Reply Quote
Kaddaman

Send message
Joined: 13 Oct 21
Posts: 7
Credit: 215,774
RAC: 1,120
Message 22857 - Posted: 15 Oct 2021, 11:55:08 UTC - in response to Message 22855.  
Last modified: 15 Oct 2021, 12:05:48 UTC

15/10/2021 13:53:40 | Cosmology@Home | Starting task camb_boinc2docker_3834991_1634297593.024318_0
15/10/2021 13:53:49 | | Vbox app stderr indicates CPU VM extensions disabled
15/10/2021 13:53:49 | Cosmology@Home | Computation for task camb_boinc2docker_3834991_1634297593.024318_0 finished
15/10/2021 13:53:49 | Cosmology@Home | Output file camb_boinc2docker_3834991_1634297593.024318_0_r1837228939_0.tgz for task camb_boinc2docker_3834991_1634297593.024318_0 absent

SVM is enabled in BIOS, now the upper errors show up..

Now it even says "CPU does not have hardware virtualization support", although VirtualBox is installed and Windows Task Manager says that Virtualization is enabled.
ID: 22857 · Report as offensive     Reply Quote
maeax

Send message
Joined: 21 Dec 17
Posts: 30
Credit: 4,312,100
RAC: 26,033
Message 22858 - Posted: 15 Oct 2021, 13:37:56 UTC - in response to Message 22857.  

Hyper-V as feature must be disabled.
Your docker tasks show CPU's in use 12.
You can check your prefs in the website and set CPU's to 1 to see how it works.
In Virtualbox program -File - virtualmedien is there a yellow triangle shown.
If yes, this files must be deleted.
ID: 22858 · Report as offensive     Reply Quote
Kaddaman

Send message
Joined: 13 Oct 21
Posts: 7
Credit: 215,774
RAC: 1,120
Message 22859 - Posted: 15 Oct 2021, 13:50:56 UTC - in response to Message 22858.  
Last modified: 15 Oct 2021, 14:15:46 UTC

I am currently pool crunching, so changing the project preferences is not possible AFAIK.
I do have 12 threads with my CPU though, so that number is correct.

I disabled Hyper-V with a CMD command I found on Reddit, and now a boinc2docker task is running. Let's see if it finishes successfully.
It is camb_boinc2docker_3948592_1634304617.061179.

However, it is only barely utilizing my CPU. The cores are almost idling (but still doing something at least) and the temperature is really low. It's also not using any RAM at all.
Is that working as intended? Maybe the readouts of CPU and RAM usage (hwinfo64 and Ryzen Master) aren't 100% correct, but regarding the temperature, I guess they are quite right.

Thanks for replying so quick by the way!

EDIT: Seemingly only 1 thread of my CPU is being used at 100% somehow.

EDIT2: Computation error again...
15/10/2021 15:55:51 | Cosmology@Home | Starting task camb_boinc2docker_3948592_1634304617.061179_0
15/10/2021 15:56:09 | Cosmology@Home | Computation for task camb_boinc2docker_3947840_1634304582.021672_0 finished
15/10/2021 16:12:58 | Cosmology@Home | Aborting task camb_boinc2docker_3948592_1634304617.061179_0: exceeded elapsed time limit 1025.40 (86400.00G/84.26G)

Gonna swap projects soon if this doesn't work. Too much possible crunch time lost already with troubleshooting...
ID: 22859 · Report as offensive     Reply Quote
maeax

Send message
Joined: 21 Dec 17
Posts: 30
Credit: 4,312,100
RAC: 26,033
Message 22860 - Posted: 15 Oct 2021, 14:30:50 UTC - in response to Message 22859.  

app_config.xml helps to reduce the CPU's to a less number.
Have only in LHC this in use, maybe you find it in a thread of Cosmology.
Therefore you need some experience.
app_config.xml is placed in the Boinc-Data folder of the Project.
Step by step... and hoping you get some successful tasks with docker.
ID: 22860 · Report as offensive     Reply Quote
Kaddaman

Send message
Joined: 13 Oct 21
Posts: 7
Credit: 215,774
RAC: 1,120
Message 22861 - Posted: 15 Oct 2021, 14:57:40 UTC - in response to Message 22860.  

Thanks for trying!

I managed to find the root cause: The docker inside the VM can't start with all 12 cores enabled. There seems to be a misunderstanding inside the VM what threads and actual cores are. I started a WU for it to set up the VM, then exited BOINC so the VM shut down, and then edited the VM to use 6 cores. Now it actually finishes the WU successfully, but all in all, it's only using 50% of my computing power. Well, since multithreading doesn't double the computing power, it uses more than 50%, but you know what I mean.
ID: 22861 · Report as offensive     Reply Quote
Rymorea

Send message
Joined: 26 Sep 15
Posts: 7
Credit: 503,903
RAC: 331
Message 22862 - Posted: 15 Oct 2021, 15:51:38 UTC

Try and error I find optimum 4 core per wu. So at preferences set Max # jobs infinite - Max # CPUs 4 cause I have 8 cores 16 treat wus use 8 core others for win and me :) Also add app_config.xml for restric only 2 wu run and other waiting start like this
<app_config>
<app>
<name>camb_boinc2docker</name>
<max_concurrent>2</max_concurrent>
</app>
<app_version>
<app_name>camb_boinc2docker</app_name>
<plan_class>vbox64_mt</plan_class>
<avg_ncpus>4</avg_ncpus>
</app_version>
</app_config>
and finally at boinc manager let them go 80 wus downlod cause cosmology server limit 80. Till get error from Virtualbox.
ID: 22862 · Report as offensive     Reply Quote
Greger

Send message
Joined: 30 Oct 15
Posts: 8
Credit: 47,535,588
RAC: 255,883
Message 22863 - Posted: 15 Oct 2021, 16:08:14 UTC

For this issue that multi threaded tasks is not using 100% or close to it is because how it made to work.

On each mt task it got a main job and each thread got each own job and these could require different runtime until finished so jobs that are done is collected in main process and threads would be in waiting for other jobs to be finished. When all is done main process would give out new jobs to all. Then it repeating this until task is fully completed. A good mt task would have jobs that is close each other but some could have one job much longer and that would hold other threads.

For task in vm-environment it would also need startup time to build programs and run script take some time and also longer to close it. So start/stop is longer then then it would be in native state or local install.

I suggest reduce threads as you did to 4-8 and reduce amount of task running concurrently at these task require more in I/O but would have lower cpu usage.
Could mix docker with legacy and run both together to balance load on system.
ID: 22863 · Report as offensive     Reply Quote
Kaddaman

Send message
Joined: 13 Oct 21
Posts: 7
Credit: 215,774
RAC: 1,120
Message 22864 - Posted: 16 Oct 2021, 3:41:15 UTC - in response to Message 22863.  

Thanks for your answers! I created an app-config.xml and set avg_ncpus to 6 and max_concurrent to 2, which works and now uses all of my 12 CPU threads.

Now the only issue that is left is that my PC becomes totally unresponsive when running a boinc2docker WU, even when only running 1 WU and lowering the avg_ncpus to 2. Reading the guides and FAQs, that was somehow expected though. I guess the only workaround is to lower the avg_ncpus while I'm using the PC and raise it again when I'm not.
ID: 22864 · Report as offensive     Reply Quote

Forums : Technical Support : camb_boinc2docker WUs failing