I thought that these longer tasks were supposed to end at 48 hours? I don't mind letting it go but I am curious if there is a problem.

i7-2600 @ 3.4 Ghz

I suspended the task and rebooted the Windows 7 PC (uptime has been a few months). When I launched BOINC they cleared and either errored or were invalid due to "Completed, too late to validate".

Tom I still get some tasks that get stuck at 100%, Win 7 64-bit.

Should we just abort them if they run > 1 hour? Most tasks complete in 15-25 minutes on a Q6600.

I've any never ending tasks.I abort it.
The normal runtime on my comp. is less than an half hour.
Greetings.

I've any never ending tasks.I abort it.
The normal runtime on my comp. is less than an half hour.
Greetings.

I discovered that with 100% repeatability, if the task is suspended and restarts for ANY reason on my systems, with keep tasks in memory NOT SELECTED, then the tasks will go into overtime while using ZERO CPU time and never finish properly. I've proved this on 3 systems. I have NOT tested this with Keep Suspended tasks in memory option set.

A Boinc Stop/Start when all the other tasks are finished will then cause the task to report, but likely get some error. Also, any other project that bumps/suspends an active mindmodel task causes the same problem... ANY reason I can find that the task starts/stops/starts causes the problem; this includes running CPU benchmarks...

Naturally, stopping/starting boinc once the task has started causes the same problem on ALL tasks already started. However, suspending UN-STARTED tasks and letting the active tasks finish before rebooting or restarting BOINC works fine...


:)

PS: Sometimes they do it when other project tasks start WITHOUT suspending an MM task... soooo, something strange in exit land on the MM tasks...

I had one task go for 1.5 hrs at 100% done and zero CPU's used, suspended the waiting new tasks, restarted BOINC, and now that task reported invalid signature... same as he others.


Win7-Compaq

51 MindModeling@Beta 7/11/2012 5:44:03 PM Restarting task MindModeling-175-4ffc4ade1b69a_0 using ccl_wrap version 175 (sse2) in slot 2
52 MindModeling@Beta 7/11/2012 5:44:16 PM Computation for task MindModeling-175-4ffc4ade1b69a_0 finished
53 MindModeling@Beta 7/11/2012 5:44:25 PM Started upload of MindModeling-175-4ffc4ade1b69a_0_0
54 MindModeling@Beta 7/11/2012 5:44:27 PM [error] Error reported by file upload server: invalid signature

I discovered that with 100% repeatability, if the task is suspended and restarts for ANY reason on my systems, with keep tasks in memory NOT SELECTED, then the tasks will go into overtime while using ZERO CPU time and never finish properly.

I don't mind setting the flag to keep things in memory except when I am running tasks that use over 2Gig Mem each.

Other tasks I run use 1.2Gig Mem each and with a 6 core CPU, one runs out of ram quickly... and windows real sucks when it starts thrashing memory off the virtual drive...

Sigh... so, in most case, I just leave it off...

Nevertheless, this is a Beta project... so things like checkpoint/save state file bugs are to be expected...

No worries... I'm sure the developers will fix it.

:)

FYI
I am seeing numerous WU's stuck at various % completion for 8.5+ hours and rising, in the current run.

Occurring on Win7-64 and Linux-64, intel and AMD
BOINC 7.0.28 (current stable)

I have aborted the WUs now as the current advice and observed norm has been 20 mins to 1 hour on these machines.

Do we have advice that the current run has moved away from the norm and to expect much longer WU's?

As an FYI, a number of the 191 tasks are also experiencing long run times > 1 hour. I've been aborting anything that goes over 1 hour run time and with the remaining time still increasing (but it's usually over an hour as well). I've got a few that I'm going to let run on my Win7 Q6600 machine just to see if they ever do complete or if they were like the earlier tasks and go 4+ hours or more.

As an FYI, a number of the 191 tasks are also experiencing long run times > 1 hour

I have a couple of 191's that went high priority after lots of hours. I did not write down exactly how many. I found this thread and shut down boinc to restart it. Boinc exited but the worker threads were still running according to ps, so I rebooted the machine.

Both work units restarted from 0 on reboot and are currently running high priority. I will give them an hour or two then abort them.

A couple of the ones I had on the Q6600 finished in about 2 hours. I have one though that is at 3:09 elapsed and 2:45 remaining (h:m) with both increasing at about the same rate. Going to abort that one.

On the ones that finished, the remaining time kept increasing and then suddenly drop 15-30 minutes, then start increasing again. As long as it falls off a cliff farther than it climbs back up before falling again, it eventually finishes.

Edit:
I also found a number of the 1.77*.exe applications left running in the background with a number of them still consuming CPU. Checking the boincmgr event log, it's filled with "app reporting negative CPU:-1.000000" error messages. Presumably this is because all those backgroud running tasks were taking CPU as well as the ones BOINC thought it was managing. So the amount of CPU actually available to BOINC was zero (all consumed by stuck apps). But windows just happily keeps scheduling them to run. Probably the same on Linux as well.

Maybe this explains the longer run times as well since the elapsed and remaining times kept increasing at 1 second intervals, but clearly, it can't get 1 second of actual work done in that timeframe due to the other apps also consuming CPU.

Just as a test, I aborted a long running task on Linux and it leaves behind the extra 1.77 process. So I now have 5 tasks - ah, it's a dual core - running. Since I've had to abort a few tasks on this machine, I suspect these have just been building up. I had to do that more on the Windows machine, so the problem was actually worse there.

I have a couple of 191's that went high priority after lots of hours. I did not write down exactly how many. I found this thread and shut down boinc to restart it. Boinc exited but the worker threads were still running according to ps, so I rebooted the machine.

Both work units restarted from 0 on reboot and are currently running high priority. I will give them an hour or two then abort them.

I have a couple of 191's that went high priority after lots of hours. I did not write down exactly how many. I found this thread and shut down boinc to restart it. Boinc exited but the worker threads were still running according to ps, so I rebooted the machine.

Both work units restarted from 0 on reboot and are currently running high priority. I will give them an hour or two then abort them.

Both units finished after a system reboot that caused them to start from scratch. They both finished in under an hour runtime. One of them previously ran for over ten hours and was using 1G of RAM and was hung at 75%. Both of the units had been aborted three other times.

edit: I should probably mention this is on a 64 bit linux machine.

There a still quite a few tasks that have very long run times. I suspect in the morning I will have most cores stuck running these and wind up having to kill them.