Long WU wont finish by deadline

[scalextrix]

Hi I have a cmsearch VM task just started 1h 48m ago, and it reports a runtime of 724h 11m at the moment, based on the current progress of 0.148% it looks like this will take 30-40 days to complete but the deadline is Dec 20th so its not clear Im going to make it in time.

I have an Intel 3770k overclocked to 4.1Ghz, HT is on so I can run 8 threads and I also have GPU tasks that use antywhere between 0.6 to 1 HT CPU core. I have 16GB RAM and run WIn 7 64Bit.

I suspended all other projects and the system CPU usage was around 14-16%, is there a way to speed this up?

[ChristianB]

That's a normal behavior for cmsearch VM tasks (formerly known as XXL). The deadline will be automatically extended on the server if the task is processing (every 4 cpu-hours a short trickle up message is send to the server). Sadly the deadline on the client won't be extended. You can check via the Website what the new deadline is. The VM will only use 1 CPU core and up to 4 GB memory so you can run other projects with this host. You can't speed the processing of the VM up, it takes what it takes. A snapshot (checkpoint) is done according to your BOINC setting "write to disk at most every x secs" so no computation is lost when you power down your computer.

[scalextrix]

Ok thanks, in any case I turned off my machine overnight, in restarting this morning it ran the WU for another 2 minutes then hit a computation error.

[Michael H.W. Weber]

Ok thanks, in any case I turned off my machine overnight, in restarting this morning it ran the WU for another 2 minutes then hit a computation error.

...indicating an error that was fixed quite some time ago if your are using the latest BOINC manager, the latest VirtualBox version and the latest VM apps.

Michael.

[scalextrix]

Ok so Im on Boinc 7.2.31 and Virtualbox 4.2.16

I know a later Virtualbox is available but Im also running test4theory and that is not allowed on a later version yet, will I have to choose which project to run based on the VM version?

I have a new RNA task now and its XXXXL 1460hours! The way Im running my machine that will probably take 4 months, is it really worth me even trying this project? I want to help but Im not sure if the science can wait for me that long (to return 1 result).

[Jacob Klein]

Ok thanks, in any case I turned off my machine overnight, in restarting this morning it ran the WU for another 2 minutes then hit a computation error.

...indicating an error that was fixed quite some time ago if your are using the latest BOINC manager, the latest VirtualBox version and the latest VM apps.

Michael.

Michael:

Can you provide more information about the error?

I looked up the user's tasks, and found:
http://www.rnaworld.de/rnaworld/result. ... d=14921969
http://www.rnaworld.de/rnaworld/result. ... d=14922386
... and I'm generally familiar with some of the errors that happen, but I'm not familiar with this error:
-2135228409 (0xffffffff80bb0007) Unknown error number
Error in start VM for VM: -2135228409

What causes it, and what fixes/fixed it?


scalextrix:

Stick with VirtualBox 4.2.16, since you are running Test4Theory. You should be able to complete RNA World tasks also, using 4.2.16. Test4Theory has issues with all later versions, including the current 4.3.4. And I would actually recommend attempting any of the RNA World VM tasks, even if they'll take a couple months. Their Monster Hunters page shows that some tasks just take lots of time, and even if your attempt isn't successful, it still helps us work out any remaining bugs. It's up to you, but don't feel like it won't be worth it.

[Michael H.W. Weber]

Ok so Im on Boinc 7.2.31

...which is outdated, too.

@Jacob: No, unfortunately I cannot tell you much about the specific error code - but the phenomenon described reminded me of the bugs which Christian has solved (he will probably also know the details of these error codes).

Michael.

[scalextrix]

I just saw the new BOINC version 7.2.33 so updated to that.

If my WU failures are helping the project that is fine, will keep running the project, don't care about scores or leader-boards, Im doing it for other reasons

Thanks to everyone on the forum, good luck.

[scalextrix]

Ok I noticed in the VirtualBox that all of the failed tasks Im getting are leaving an Inaccessible image (error -103).

I do also use the VirtualBox to run other guest OS's, for example Im looking around Win Server 2012 on an evaluation copy right now, and Im looking at PuppyLinux as well. Im vaguely aware that the WU failures may coincide with me starting/stopping other guest OS's, does that sound probable? Perhaps Im just making connections that arent there.

[ChristianB]

Ok I noticed in the VirtualBox that all of the failed tasks Im getting are leaving an Inaccessible image (error -103).

I do also use the VirtualBox to run other guest OS's, for example Im looking around Win Server 2012 on an evaluation copy right now, and Im looking at PuppyLinux as well. Im vaguely aware that the WU failures may coincide with me starting/stopping other guest OS's, does that sound probable? Perhaps Im just making connections that arent there.

We are not sure but it generally shouldn't impact our computation if you run other VM. The inaccessible entries in VBoxManager are new and may clear up after a restart but shouldn't impact the usage of VirtualBox at all.

[scalextrix]

So Im pleased to report a WU with CPU time over 25 hours now, the Application name has changed, looking at the task history all the ones that failed were:

cmsearch VM (VirtualBox) 1.0.2 1.06 (vbox64)

Whereas the one I seem to be crunching well is:

cmsearch VM (VirtualBox) 1.0.2 1.08 (vbox64)

Hope that helps someone.

[Michael H.W. Weber]

Yes, we updated the client software.

Michael.