VBscript Fights Cuda

[Jason G]

I finished with AP testing, and loaded up the 6.08 cuda app with VLAR Kill mod, and still have been getting frozen tasks, So I've just now enabled the script again.

Since they fixed a lot of things, and the app itself will be terminating the Sluggish VLAR's anyway, one thing that seems to work for me is restarting Boinc.  Is it possible to modify the script such that it restarts Boinc on detection of a stuck task,(with perhaps a few seconds in between stopping then starting the service... for us with service install anyway  ) rather than terminating the task?

Anyway, seeing how it will go.  Pretty good if I can stop having to manually check the machine every few minutes  .

[Jason G]

I've had one stick so far that the script seemed to not be able to terminate?

Relevant Log fragment: (Sorry for long post  ) [Edit: Not happened again since this one time, a fluke ?]

1/26/2009 1:25:48 PM > -watching new task: 16dc08ad.22822.1708.5.8.113_1
1/26/2009 1:25:48 PM > -true angle range: 0.43389603140378
1/26/2009 1:25:48 PM >
1/26/2009 1:25:48 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (1. time)
1/26/2009 1:26:34 PM > ID: 2260 0%, running: 46s
1/26/2009 1:26:34 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (2. time)
1/26/2009 1:27:19 PM > ID: 2260 0%, running: 91s
1/26/2009 1:27:19 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (3. time)
1/26/2009 1:28:05 PM > ID: 2260 0%, running: 137s
1/26/2009 1:28:05 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (4. time)
1/26/2009 1:28:05 PM >
1/26/2009 1:28:05 PM > ! terminating process MB_6.08_mod_VLAR_kill_CUDA.exe
1/26/2009 1:28:05 PM > ! - crashed WU-file: 16dc08ad.22822.1708.5.8.113_1
1/26/2009 1:28:05 PM >  --- > RunningTime: 137 --- MinRunTime: 120
1/26/2009 1:28:05 PM >
1/26/2009 1:28:05 PM > trying to copy wu
1/26/2009 1:28:05 PM > File: 16dc08ad.22822.1708.5.8.113.wu copied ...
1/26/2009 1:28:05 PM > --> trying copy of stderr ...
1/26/2009 1:28:05 PM > --> WU_copy_log.txt found
1/26/2009 1:28:05 PM > --> client_state.xml found
1/26/2009 1:28:50 PM > ID: 2260 0%, running: 182s
1/26/2009 1:28:50 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (1. time)
1/26/2009 1:29:36 PM > ID: 2260 0%, running: 228s
1/26/2009 1:29:36 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (2. time)
1/26/2009 1:30:21 PM > ID: 2260 0%, running: 273s
1/26/2009 1:30:21 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (3. time)
1/26/2009 1:31:07 PM > ID: 2260 0%, running: 319s
1/26/2009 1:31:07 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (4. time)
1/26/2009 1:31:07 PM >
1/26/2009 1:31:07 PM > ! terminating process MB_6.08_mod_VLAR_kill_CUDA.exe
1/26/2009 1:31:07 PM > ! - crashed WU-file: 16dc08ad.22822.1708.5.8.113_1
1/26/2009 1:31:07 PM >  --- > RunningTime: 319 --- MinRunTime: 120
1/26/2009 1:31:07 PM >
1/26/2009 1:31:07 PM > trying to copy wu
1/26/2009 1:31:07 PM > 16dc08ad.22822.1708.5.8.113.wu already exists, no copy ne
eded
1/26/2009 1:31:07 PM > --> trying copy of stderr ...
1/26/2009 1:31:07 PM > --> WU_copy_log.txt found
1/26/2009 1:31:07 PM > --> client_state.xml found
1/26/2009 1:31:52 PM > ID: 2260 0%, running: 364s
1/26/2009 1:31:52 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (1. time)
1/26/2009 1:32:38 PM > ID: 2260 0%, running: 410s
1/26/2009 1:32:38 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (2. time)
1/26/2009 1:33:23 PM > ID: 2260 0%, running: 455s
1/26/2009 1:33:23 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (3. time)
1/26/2009 1:34:09 PM > ID: 2260 0%, running: 501s
1/26/2009 1:34:09 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (4. time)
1/26/2009 1:34:09 PM >
1/26/2009 1:34:09 PM > ! terminating process MB_6.08_mod_VLAR_kill_CUDA.exe
1/26/2009 1:34:09 PM > ! - crashed WU-file: 16dc08ad.22822.1708.5.8.113_1
1/26/2009 1:34:09 PM >  --- > RunningTime: 501 --- MinRunTime: 120
1/26/2009 1:34:09 PM >
1/26/2009 1:34:09 PM > trying to copy wu
1/26/2009 1:34:09 PM > 16dc08ad.22822.1708.5.8.113.wu already exists, no copy ne
eded

 Then I manually aborted the task (actually killed it it task manager),  Though a restart of the Boinc service would have allowed it to process, more than likely:

1/26/2009 1:34:09 PM > --> trying copy of stderr ...
1/26/2009 1:34:09 PM > --> WU_copy_log.txt found
1/26/2009 1:34:09 PM > --> client_state.xml found
1/26/2009 1:34:54 PM > ID: 2260 0%, running: 546s
1/26/2009 1:34:54 PM > MB_6.08_mod_VLAR_kill_CUDA.exe at 0% (1. time)
1/26/2009 1:35:40 PM >
1/26/2009 1:35:40 PM > -watching new task: 16dc08ad.22822.1708.5.8.74_1
1/26/2009 1:35:40 PM > -true angle range: 0.43389603140378
1/26/2009 1:35:40 PM >
1/26/2009 1:35:40 PM > resetting counter
1/26/2009 1:35:40 PM > -increasing BreakPerCycle to 43 sec
1/26/2009 1:35:40 PM > -increasing CpuReadTime to 4 sec
1/26/2009 1:36:27 PM > ID: 1864 11%, running: 47s

[Maik]

@Jason: I cant reconstruct the error, sorry.
I see the script wasn't termianting the task because of the process-ID did not change.
Its the same error like RandyC reported.

@RandyC: What OS are you using? The 'VB-Script terminate command' does not work on all OS.
And no. You don not need -sam to make the script terminating idleing taks processes.

Maybe booth could attach a full (not snipped) log.txt next time?
I wasn't really thinking that the script is needed such a long time so I did not add any routine to print debug informations with internal variable settings ^^

[Jason G]

Sure, will attach a full log if it happens again.  Have run out of MB work since then, and it never happened again.

[randyconk]

@Jason: I cant reconstruct the error, sorry.
I see the script wasn't termianting the task because of the process-ID did not change.
Its the same error like RandyC reported.

@RandyC: What OS are you using? The 'VB-Script terminate command' does not work on all OS.
And no. You don not need -sam to make the script terminating idleing taks processes.

Maybe booth could attach a full (not snipped) log.txt next time?
I wasn't really thinking that the script is needed such a long time so I did not add any routine to print debug informations with internal variable settings ^^

I am using XP Pro 32b SP3. I don't know if that's my problem or not. One thing I thought of a while ago: I am using a Service install for BOINC. It may be that with a Service install, the userid running your script is not allowed to terminate BOINC's tasks.

[Jason G]

I am using XP Pro 32b SP3. I don't know if that's my problem or not. One thing I thought of a while ago: I am using a Service install for BOINC. It may be that with a Service install, the userid running your script is not allowed to terminate BOINC's tasks.

exact same scenario here(XPsp3 & Service), but the script should be running as 'me' right? so in my case has admin privileges AFAIK.

[randyconk]

I am using XP Pro 32b SP3. I don't know if that's my problem or not. One thing I thought of a while ago: I am using a Service install for BOINC. It may be that with a Service install, the userid running your script is not allowed to terminate BOINC's tasks.

exact same scenario here(XPsp3 & Service), but the script should be running as 'me' right? so in my case has admin privileges AFAIK.

Not sure about that (my id has admin privileges too)...BOINC 6.x.x (running 6.4.5) sets up a special set of IDs to run Service installs. Could try starting the script using 'run-as' option and see if that fixes it. I'd need to set-up a shortcut to start the script.

[Maik]

Hehe, process terminat works at XP. Im using that too. but not installed as Service. Maybe thats the reason ...

[randyconk]

Hehe, process terminat works at XP. Im using that too. but not installed as Service. Maybe thats the reason ...

Can you give me a code segment (and tell me where to insert it) that will perform the following (I don't code VBS) in place of the terminate:
  net stop boinc
  wait 30 seconds
  net start boinc

[Maik]

Not atm, sorry.
I have no time atm. Maybe in 3 or 4 days.

[randyconk]

Not atm, sorry.
I have no time atm. Maybe in 3 or 4 days.

OK. That was going to take too long, so I did some surfing and found a VBS and wscript tutorial and spent about 3 hours on it last night working up some code. It's kind of brute force and could be accomplished better if I knew VBS and wscript more, but I think it will work. See attached for a write-up of what I did.

[attachment deleted by admin]

[randyconk]

Not atm, sorry.
I have no time atm. Maybe in 3 or 4 days.

OK. That was going to take too long, so I did some surfing and found a VBS and wscript tutorial and spent about 3 hours on it last night working up some code. It's kind of brute force and could be accomplished better if I knew VBS and wscript more, but I think it will work. See attached for a write-up of what I did.

I'm home from work now and I checked that system. It showed 3 stalled WUs and the script restarted BOINC each time. I think we have a winner.

If you can come up with something more efficient, that'd be neat too.

[Jason G]

Batch restart service mod works well!. Two days now & it only needed to restart once, but it did so and resumed processing on GPU normally.

Jason

[Raistmer]

Is it possible to watch for 2 tasks, not one? (for dual GPU setup)

[chelski]

[Edited] - Sorry please ignore the post, somehow manage to get the script running again, must be something wrong at my side.