VBscript Fights Cuda

[Maik]

Hi there!

Maybe you know me from Seti Board. Raistmer was suggesting me to join here.
Here i am ! You should know my first language is German, so please be patient with me if my english looks a bit silly ^^
Ok, on work:

Maybe you know i have written a VBscript that watches Cuda-app.exe and fights with its freezing.

vbscript is modified and will copy automaticly wu's which get stuck, tested -> works
now ill try to modify the script that it also will do a copy of WU's that ran to short (compute error) ...
then i'll have my full-auto-watching-wu-and-copy-it-to-stand-alone-dir-vbscript )

Wow, very nice !
It seems it's very useful to master VB scripting, will look in this direction, thanks for info

After that PM i was thinking i should report here what i have done:
 # i created 2 scripts, one for normal use (cuda_log), one for users who do stand alone tests (cuda_log_test)
 -> cuda_log activity:
 ' watching cuda-app.exe
 ' terminating it when cpu-usage is 0 over 60sec
 ' logging to file / monitoring at cmd-window all activities
 -> cuda_log_test activity:
 ' same as cuda_log +
 ' counting task-runtime and monitoring / logging it
 ' if runtime below defined time -> copy WU to defined folder (to catch compute errors / result overflows)
 ' if task get freezing -> copy WU to defined folder
 ' (on my to-do-list -> copy of stderr)

Be carful if you use that 'automated' WU-cleaning (process-terminating). Check your host from time to time.
Some WU's are causing your grafic driver to crash. The result are a lot of failing WU's in a row (noticed that this morning at my host).
The crash announces itself with around jumping pixels everywhere. Some users posted that already ...

If you want to use the script please focus your attention to the settings. Wrong settings will make the script to fail.
I will explain them in my post.
The scripts is running fine at my host (WinXP Pro SP3).
If the script want start, maybe 'scipting host' is deactivated (via xp-AntiSpy or other tools like this)
Edit: delete files-downloads (newer one)
Extract the file where you want and where you have full access (create-file, change-file, e.g. C:\temp).
DONT double-click the *.vbs-file. Windows will use the wrong application tu run the script. Just double-click the *.cmd-file, its the starter-file for the *.vbs-files.

Example from script output:

04.01.2009 01:48:11 > Process-Control for MB_6.06r380mod_CUDA.exe
04.01.2009 01:48:11 > logFile: pc_log_0_04012009.txt
04.01.2009 01:48:11 > maxLogFileSize set to: 51200 bytes
04.01.2009 01:48:11 > boinc dataDir found
04.01.2009 01:48:11 > standAlone folder found ... sub-standAlone activated
04.01.2009 01:48:11 > - running ----------------------------------------
04.01.2009 01:48:11 > 
04.01.2009 01:48:14 >  --- watching new WU-file: 06no08aa.9787.12342.9.8.106_0
04.01.2009 01:48:29 > ID: 3540 MB_6.06r380mod_CUDA.exe 6 %, running: 15s
...
04.01.2009 01:55:01 > ID: 3540 MB_6.06r380mod_CUDA.exe 12 %, running: 407s
04.01.2009 01:55:16 >  --- watching new WU-file: 06no08aa.9787.12342.9.8.125_0
04.01.2009 01:55:31 > ID: 2020 MB_6.06r380mod_CUDA.exe 36 %, running: 15s
...
04.01.2009 02:03:32 > ID: 2020 MB_6.06r380mod_CUDA.exe 12 %, running: 496s
04.01.2009 02:03:47 > ID: 2020 MB_6.06r380mod_CUDA.exe 12 %, running: 511s
04.01.2009 02:04:02 >  --- watching new WU-file: 06no08aa.9787.12342.9.8.93_1
04.01.2009 02:04:18 > ID: 2988 MB_6.06r380mod_CUDA.exe 18 %, running: 16s
04.01.2009 02:04:33 >  --- watching new WU-file: 06no08aa.9787.12342.9.8.112_0
04.01.2009 02:04:33 > 
04.01.2009 02:04:33 >  --- debug > RunningTime: 31 --- MinRunTime: 120
04.01.2009 02:04:33 > File: 06no08aa.9787.12342.9.8.93.wu copied ...
04.01.2009 02:04:33 > 
04.01.2009 02:04:33 > 
04.01.2009 02:04:48 > ID: 1016 MB_6.06r380mod_CUDA.exe 4 %, running: 15s

[Maik]

To change settings right-click the *.vbs-file -> context-menue comes up -> choose open with ... -> use notepad.exe
All settings are explained with comments (comments beginning with ')
Settings in cuda_log are easy to understand i think.

For cuda_log_test users, scroll a bit down, there you will find a comment
 'settings only for stand alone to copy WU's

To make the script to copy files ssp must be true (ssp = true)
To set your temp-folder is sspp for ...
sspp = "D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\"
Here set the time in seconds that the WU should run to be not a cumpute error / result overflow
(normally WU's are canceled from set-app after 30 sec if result overflow happened)
appMinRunTime = 120

OK, nuff said ... Your turn

[Jason G]

Sounds like I might be able to try the App again with this, thanks Maik, Will try later today.  May take a looksie what is causing these issues If I get a chance. 

[Maik]

-
I have a request:
If you use the script please notify me about errors (with error messages).
I cant bugfix it if i read posts like 'i have changed that and now it runs' as it happend in seti-forum.
-
Thanks

[Raistmer]

@Maik - can't send PM to you - PM limit reached. I deleted some messages but still the same - maybe it's your PM limit reached not my ?
Check this, please.

[Maik]

' (on my to-do-list -> copy of stderr)

I have finished that little project now. Some tests on my host ended sucessfull.
The zip-file contains: *cuda_log_030b.cmd, *cuda_log_030b.vbs, *readme_030b.txt

Please report back on a sucessfull / unsucessfull use of it.

EDIT:
Notice: This script is written for users who want to have a copy of crashed WU's (Stand-Alone-Tester)
For 'normal' users please use cuda_log.zip posted above.

EDIT: deleted file - newer one

[Maik]

 > Update 040beta
 ================
 > I've re-written the cuda_log_030b.vbs-code.
 > It includes now both versions of my basic scripts.
 > The script will now be started with options.
 > Also i have included tests for file-handling / specified folders.
 > Addon: a cuda_log.ini with all settings to handle settings better.
 > sart with option -ver and/or -sam or -non
 > -ver will show all informatiosn to cmd-window (without the script is only logging to log-file)
 > -sam will copy crashed WU's to a specified folder and will try to get a stderr from client_state.xml
 > -non will disable both options above
 >
 > to change start-options edit the cuda_log_040b.cmd - file and change it there
 > or run manual start from cmd-window while using the "cscript cuda_log_040b.vbs -option" command

Edit:
I wasnt able to do a complete test with -sam option because i dont have stucking WU's the last 24h ^^
Edit2:
Had a stucking WU now. -sam options seem to wotk without errors.

End at the END: Please report back how the script works for you.

Edit3: Once again. If you use my Script dont leave your PC alone.
If you notice a lot of failing Tasks in a row stop seti and reboot your PC.

[Maik]

minor update that fix wrong variables needed with -sam option to detect result overflows.
my fault, sorry for that ^^

It is a bit tricky to handle with variables when VBscript only know the variant-typ....


notice: the script seems to work correctly now. maybe ill rename it to v1.0 if i get no response with errors

External link for download

[popandbob]

Tried the new version... Here is the log... (Yes that is my Boinc DATA Directory... )

starting script ...
1/10/2009 12:52:09 PM > ============================================
1/10/2009 12:52:09 PM > = WU / Cuda watch script by Maik v0.41beta =
1/10/2009 12:52:09 PM > =    maybe not working on Win 95/98/NT     =
1/10/2009 12:52:09 PM > ============================================
1/10/2009 12:52:09 PM > =    option on: print to cmd-window        =
1/10/2009 12:52:09 PM > =    option on: copy crashed WU            =
1/10/2009 12:52:09 PM > ============================================
1/10/2009 12:52:09 PM > 
1/10/2009 12:52:09 PM > ini: appName > MB_r396mod_CUDA.exe
1/10/2009 12:52:09 PM > ini: logFileName > cuda_log
1/10/2009 12:52:09 PM > ini: maxLogFileSize > 50000
1/10/2009 12:52:09 PM > ini: setiPath > C:\Program Files\BOINC\projects
1/10/2009 12:52:09 PM > ini: sspp > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\
1/10/2009 12:52:09 PM > ini: appMinRunTime > 120
1/10/2009 12:52:09 PM > ini: showId > 1
1/10/2009 12:52:09 PM > ini: showName > 0
1/10/2009 12:52:09 PM > ini: showUsage > 1
1/10/2009 12:52:09 PM > ini: showRuntime > 1
1/10/2009 12:52:09 PM > Folder valid:
1/10/2009 12:52:09 PM > C:\Program Files\BOINC\projects
1/10/2009 12:52:09 PM > debug:
1/10/2009 12:52:09 PM > invalid Folder!
1/10/2009 12:52:09 PM > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\
1/10/2009 12:52:09 PM > no stand alone directory set, process/WU-watching only
1/10/2009 12:52:09 PM > log-file handling successful
1/10/2009 12:52:09 PM > wu-file-handling turned off
end of startup ...
1/10/2009 12:52:09 PM > logFile: cuda_log_0_1102009.txt
1/10/2009 12:52:09 PM > maxLogFileSize set to: 50000 bytes
1/10/2009 12:52:09 PM > --running-----------------
1/10/2009 12:52:12 PM > 
1/10/2009 12:52:12 PM > -watching new WU-file:
1/10/2009 12:52:27 PM > ID: 2688 16%, running: 15s
1/10/2009 12:52:43 PM > ID: 2688 14%, running: 31s

Reports invalid folder but it still monitors wu's....

[Maik]

it should look like that:

10.01.2009 17:37:00 > ini: appName > MB_r396mod_CUDA.exe
10.01.2009 17:37:00 > ini: logFileName > cuda_log
10.01.2009 17:37:00 > ini: maxLogFileSize > 50000
10.01.2009 17:37:00 > ini: setiPath > D:\boinc_data\projects\setiathome.berkeley.edu\
10.01.2009 17:37:00 > ini: sspp > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\
10.01.2009 17:37:00 > ini: appMinRunTime > 120
10.01.2009 17:37:00 > ini: showId > 1
10.01.2009 17:37:00 > ini: showName > 1
10.01.2009 17:37:00 > ini: showUsage > 1
10.01.2009 17:37:00 > ini: showRuntime > 1
10.01.2009 17:37:00 > Folder valid:
10.01.2009 17:37:00 > D:\boinc_data\projects\setiathome.berkeley.edu\
10.01.2009 17:37:00 > Folder valid:
10.01.2009 17:37:00 > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\
10.01.2009 17:37:00 > log-file handling successful
10.01.2009 17:37:00 > wu-file-handling successful
end of startup ...
10.01.2009 17:37:00 > logFile: cuda_log_0_10012009.txt
10.01.2009 17:37:00 > maxLogFileSize set to: 50000 bytes
10.01.2009 17:37:00 > --running-----------------
10.01.2009 17:37:03 > 
10.01.2009 17:37:03 > -watching new WU-file: 06no08ab.14392.175545.9.8.162_0
10.01.2009 17:37:18 > ID: 344 MB_r396mod_CUDA.exe 4%, running: 15s
10.01.2009 17:37:33 > ID: 344 MB_r396mod_CUDA.exe 6%, running: 30s

The 'valid/invalid' - message tells you that this folder exists.
At your startup the seti folder is missing:
 >> 1/10/2009 12:52:09 PM > C:\Program Files\BOINC\projects  .... <-----
Also the WU-Name is missing (that tells yout the setiPath is invalid):
 >> 1/10/2009 12:52:12 PM > -watching new WU-file: .... <-----

The Script is only monitoring the *.exe-file set in *.ini at your WinTaskManager.
It will notice a change to new wu-file only because the Process-ID will change if a new task is starting.

- Additional: Delete the 'D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\' - entry in the *.ini-file if you dont need copy's of crashed wu's or start without -sam

Edit: I see, its my fault. The *.ini-information is wrong. It should be:
  set seti's data directory , e.g. C:\ProgramData\BOINC\projects\setiathome.berkeley.edu\

[popandbob]

Ah I see I miss read the log... The valid/invalid line is before it says what the folder is...

1/10/2009 12:52:09 PM > Folder valid:
1/10/2009 12:52:09 PM > C:\Program Files\BOINC\projects


1/10/2009 12:52:09 PM > debug:
1/10/2009 12:52:09 PM > invalid Folder!
1/10/2009 12:52:09 PM > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\

So it has the folder for the first but not the stand alone.

[Maik]

Ah I see I miss read the log... The valid/invalid line is before it says what the folder is...

1/10/2009 12:52:09 PM > Folder valid:
1/10/2009 12:52:09 PM > C:\Program Files\BOINC\projects

It just tells you, that the folder exists on your drive.
If your setup is valid then you will see the WU/Task-name in script output.

I'm working on 042b. I'll add better output-messages

[Maik]

v042 , have fun

better startup messages now:

11.01.2009 20:52:22 > testing folder ...
11.01.2009 20:52:22 > D:\boinc_data\projects\setiathome.berkeley.edu\
11.01.2009 20:52:22 > ... folder found
11.01.2009 20:52:22 > 
11.01.2009 20:52:22 > testing folder ...
11.01.2009 20:52:22 > D:\boinc_data\KWSN Knabench 1.43\TestWUs\Reserve\
11.01.2009 20:52:22 > ... folder found
11.01.2009 20:52:22 > 
11.01.2009 20:52:22 > log-file handling successful
11.01.2009 20:52:22 > wu-file-handling successful

and a little bonus

11.01.2009 20:53:28 > -watching new task: 16no08aa.10396.13161.10.8.179_2
11.01.2009 20:53:28 > -true angle range: 2.7185053469689

External link for download

[Devaster]

it gives me type mistmash at ar 0.3 ... (WS2008 eng)

[Richard Haselgrove]

Maik,

Have you thought of using boinccmd to automate the VLAR removal problem? (Arises from your remark at SETI).

Look at http://boinc.berkeley.edu/wiki/Boinccmd_tool (first entry under "Control operations" - boinccmd still calls a unit of work a "result", where the rest of BOINC now calls it a "task"). 'Abort' is one of the result-level operations supported.

If you could find a way of doing that too, then we would have a total package for unattended running.