Rechenkraft.net e.V.

I just got a task, and it's in a state I've never seen before.

cmsearch VM (VirtualBox) 1.0.2 1.19 (vbox64)
cmsvm_GA-p[e20-30MB_Lin64f]_1_Ozyzias_latipes (Japanese-medaka) more not copied
State: Postponed: VM job unmanageable, restarting later.

From log file:

6/9/2021 1:31:58 AM | | Starting BOINC client version 7.16.11 for windows_x86_64
6/9/2021 1:31:59 AM | | VirtualBox version: 6.1.12

Does this mean something unexpected? For example, is the VirtualBox version incompatible with this task?

This workunit has already failed for a long list of users, so is the workunit faulty?

> State: Postponed: VM job unmanageable, restarting later.

I've found a procedure that restarts such tasks.

Shut down BOINC Restart Windows. Start BOINC.

More inconvenient than the XXL tasks, but it looks like it will work if I check often enough.

This is actually a regular XXL task.
The message reported is new to me, however.

Michael.

Robert,

1) Your discussion about your task, should have been done using a new thread, instead of hijacking the "cross validation" thread. If you'd like to continue the discussion, please start a new thread.

2) Your Task is in fact a VM task. The work-unit link is here: https://www.rnaworld.de/rnaworld/workun ... id=6330896 ... and it should complete successfully if you can keep it running long enough, with as few interruptions as possible. I should know - I already completed it, and you are my wingman on this workunit.

3) "Postponed: VM job unmanageable" happens when BOINC loses communications with the VirtualBox executables that control the VM. The exact cause is usually unknown. And yes, restarting the PC can help.

4) There is a "progress.txt" file within the slots folder for that task. If you open it to take a peek, you should hopefully see it progressing. Immediately close that text file after you've peeked, to not interfere with the task. You can keep an eye on that percentage value over time, to know that the task is working. Note that a value of 0.98765 is the highest it will go (and it will show 98.765% in the UI), but that DOES NOT MEAN that it's stuck -- It is still working! You can just look at the modified date/time of that progress.txt file, and if it's updating every half hour, it is still progressing normally.

5) It took me 228.5 days of processing time, to complete that one. I estimate it will take you about that long (since amazingly my CPU that completed it is the same model as yours) or it may take longer (since I carefully set a fully stable overclock)... so... try to keep your PC as stable as possible, and good luck!

...I split the thread accordingly.

Michael.

Thanks Michael, I appreciate that.

I've found that shutting down and restarting only BOINC also restarts such tasks but only for a few hours rather than about a day I get if I also restart Windows. This suggests that increasing clutter on the memory space is related to the cause of the problem.

You might try to increase resource share for the project to the upper limit, and adjust your Boinc preferences to switch between applications every 999 minutes.
That way other projects and the OS should not interfere too much with the RAM you need on vbox.
I also adjusted RAM settings for in use and while idle to the same amount, so it doesn't get swapped around.

It currently looks like I'll finish this task about the middle of February, if the progress percentage is accurate. Does that mean that I cannot take a Christmas vacation of more than one day, without having this task time out and fail?

The task restarted after 24 hours of being postponed, so I can probably take vacations rather than watching the task every day. Why does it take this long to restart?

I adjusted BOINC to switch between tasks every 999 minutes. The next postponed failure came much faster than that.

Increasing the percentage for RNA World will be harder - I have to find a lost password first.

I found some log files from the virtual machine. Look at the *.txt files in the slot directory.

None of them appear to contain useful information on why the virtual machine got confused, though.

An idea to check, though: After BOINC Manager sends a trickle-up message, does the server send back anything that will allow the task to drop out of the high-priority mode normally used for tasks that are too close the their deadlines or even past their deadlines?

I know that BOINC Manager doesn't display the new deadline, but does it still use the new deadlines or something derived from them to decide whether the tasks should stay in high-priority mode?