climateprediction.net home page
Posts by cooper

Posts by cooper

1) Questions and Answers : Windows : BOINC v6.10.36: Won\'t finish benchmark... (Message 39230)
Posted 14 Mar 2010 by Profile cooper
Post:
Issue fixed with v6.10.37, although not mentioned in the version history.

cooper.
2) Questions and Answers : Windows : BOINC v6.10.36: Won\'t finish benchmark... (Message 39202)
Posted 9 Mar 2010 by Profile cooper
Post:
08-Mar-2010 21:00:18 [---] Suspending computation - CPU usage is too high

There is a new preference to control this behavior...

I am aware of this.

The bench mark failure happens for me occasionally too. Normally It runs fine if I start it again manually.


This should not be necessary, at least in v6.10.32 it always recovered (v6.10.36 did not). I assume this is connected to the initial problem or the same.
I was a little surprised when I saw that v6.10.36 was released from beta to stable.

cooper.
3) Questions and Answers : Windows : BOINC v6.10.36: Won\'t finish benchmark... (Message 39197)
Posted 8 Mar 2010 by Profile cooper
Post:
Hi,

seems I am having a special issue with BOINC again...

The benchmark at startup will not finish with BOINC v6.10.36 (Win XP32). Messages:

08-Mar-2010 20:30:09 [---] Reading preferences override file
08-Mar-2010 20:30:09 [---] Preferences:
08-Mar-2010 20:30:09 [---] max memory usage when active: 1791.05MB
08-Mar-2010 20:30:09 [---] max memory usage when idle: 3223.89MB
08-Mar-2010 20:30:09 [---] max disk usage: 7.48GB
08-Mar-2010 20:30:09 [---] max CPUs used: 2
08-Mar-2010 20:30:09 [---] don\'t use GPU while active
08-Mar-2010 20:30:09 [---] suspend work if non-BOINC CPU load exceeds 25 %
08-Mar-2010 20:30:09 [---] max download rate: 40960 bytes/sec
08-Mar-2010 20:30:09 [---] max upload rate: 10240 bytes/sec
08-Mar-2010 20:30:09 [---] (to change, visit the web site of an attached project,
08-Mar-2010 20:30:09 [---] or click on Preferences)
08-Mar-2010 20:30:09 [---] Using proxy info from GUI
08-Mar-2010 20:30:09 [---] Not using a proxy
Initialization completed
08-Mar-2010 20:30:09 [---] Running CPU benchmarks
08-Mar-2010 20:30:09 [---] Suspending computation - running CPU benchmarks
08-Mar-2010 20:30:41 [---] FP benchmark ran only 1.390625 sec; ignoring
08-Mar-2010 20:47:49 [Collatz Conjecture] task collatz_1267889401_72584_0 resumed by user

After resuming tasks, the status is \'Suspended\', not \'Task suspended\'.

One the next start BOINC did this

08-Mar-2010 21:00:18 [---] Suspending computation - CPU usage is too high

and nothing else for 5 minutes. Older versions usually recover within max 1 minute.

Nothing in the changelogs, no ticket in trac, nothing in the forums?

v6.10.32 was fine. Will go back to the old version (again).

cooper.
4) Questions and Answers : Windows : Lots of model errors... (Message 39023)
Posted 25 Feb 2010 by Profile cooper
Post:

...
It is exactly the same issue, but the OP (Richard Buteau) did not come back with the content of the file, C:\\Program in his case. Interesting that his file is dated

> 02/16/2007 03:52 PM 527,671 (bytes)

but his Program Files directory is NEWER:

> 02/04/2009 09:24 AM <DIR> Program Files

How is this possible? And have a look at the size... mine only has 85 bytes.
...


Forget my question about the date of the directory.
Only the size is interesting.

cooper.
5) Questions and Answers : Windows : Lots of model errors... (Message 39022)
Posted 25 Feb 2010 by Profile cooper
Post:
Hi guys (and gyrls, if any),

with the overwhelming help of Thyme Lawn, I was able to fix it.

The reason is a single file in root directory, where BOINC runs. this may be \'C:\\Program\' or \'D:\\Program\' or \'C:\\Documents\'. In my case it was C:\\Dokumente (on a German Windows the user\'s path is \'C:\\Dokumente und Einstellungen\', as C:\\Documents and Settings on English Windows).

This file was created 27.01. at 02:13 by something out of my control, over the net, undiscovered by my antivirus software, created in the middle of the night. It contained a well known DOS message \'The command \"sh\" could not be found\' (in German). I am unable to determine what script or who was able to pass all filters (XP firewall, router filter, Windows doors etc.). In my worst dream it could have been BOINC or CPDN itself...

\'sh\' is a UNIX command. I was running a download of an linux ISO at the date of creation in that night. My antivirus was stating an error, being unable to see that server ~30 minutes earlier (before the file was created). I can\'t see why a (UNIX-)server should try to run a bash (\'sh\') script on my PC.

~

However, I was able to proof that this is the reason. After renaming the file, CPDN started crunching 2 models WITHOUT crashing the model in first seconds.

I stopped BOINC with 2 running models after 1 hour, renamed the file back to \'C:\\Dokumente\' and restarted BOINC. Both CPDN models crashed

hadsm3fub_jnsr_006441365_4
and
hadsm3fub_jnsu_006441368_7

at 22:16 my time today. A third model crashed after these 2. (Now I reached my quota.) I took a ProcessMonitor log as well.

Thyme pointed me to an older post
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=6484&nowrap=true#36114

It is exactly the same issue, but the OP (Richard Buteau) did not come back with the content of the file, C:\\Program in his case. Interesting that his file is dated

> 02/16/2007 03:52 PM 527,671 (bytes)

but his Program Files directory is NEWER:

> 02/04/2009 09:24 AM <DIR> Program Files

How is this possible? And have a look at the size... mine only has 85 bytes.



This is a stupid error. A small glitch in CPDN software seems causing models to crash; unsufficient error messages (and different!) make it hard to track it down.

Thyme passed this issue on to CPDN devs. Thanks again, Thyme!

Cheers. I\'ll have a beer. Now.

6) Questions and Answers : Windows : Lots of model errors... (Message 38981)
Posted 23 Feb 2010 by Profile cooper
Post:
What do you mean with \'how BOINC is installed\'?

Is it installed in Protected mode, (formally known as Service mode), or in Unprotected mode?



Hi Les,

not protected. I didn\'t tick that option while setup. I believe it always was like that.

HTH.
Cooper.
7) Questions and Answers : Windows : Lots of model errors... (Message 38979)
Posted 23 Feb 2010 by Profile cooper
Post:
It could be a permissions problem. Try deleting *_se_*.dll, *_se_*.exe and *_um_*.exe from projects/climateprediction.net under your BOINC data directory. They are extracted from the equivalent .zip files and their access rights might have been messed up.


oh NO...permission problem...? I always have troubles with this bloody NTFS... not only at home. Also in the office.

Backuped CPDN project directory.
Checked if I can touch all EXE and DLL. Positive. No rights problems, as I know them. (XP Home does not have the extended rights management of XP Pro).
Then deleted 6 EXEs and 2 DLLs. 2 EXE for hadsm3 6.07 and 2 EXE for hadam3p 6.14 remain. No other EXE.

Currently BOINC is downloading an *_init.gz, as usual (???!!) Shouldn\'t these be kept in the project folder? Takes a while @ 4kB/s for ~28MB...

~~~~~

OK. Pfff... finished. Here is what I did as d/l of task was completed:

-suspended all projects (in fact a few minutes before completing d/l)
-looked into 2005_12_init.gz (unpacked with 7ZIP): just normal binary stuff
-switched off av
-started SI\'s ProcessExplorer
-resume CPDN
-some EXE and DLL were unpacked by BOINC automagically
-model crashed within seconds
-report to server finished
-2005_12_init.gz was deleted by BOINC on finish of reporting
-quota of 3 results/day reached

Currently residing in project directory after report (unpacked by BOINC, previously deleted with intention):
.hadsm3_um_6.07_windows_intelx86.exe
.hadsm3_se_6.07_windows_intelx86.exe
.hadam3p_um_6.14_windows_intelx86.exe
.hadam3p_se_6.14_windows_intelx86.dll
.hadam3p_se_6.06_windows_intelx86.dll was not restored, so this is missing.

If you or someone else is interested in the ProcessExplorer logs or content of directories for debugging, let me know.

Why the hell are the *_init.gz deleted upon reporting? BOINC uses 1.47GB, free for BOINC is 8.24GB. The *_init.gz are just 28MB...

But the *_init.gz files are NOT the reason for my problems. I also can exclude my av and also permissions in BOINC directories, imho.

If that fixes the problem it would be useful to know if you\'ve changed how BOINC is installed on your system. If it doesn\'t emptying the project folder should sort things out.

What do you mean with \'how BOINC is installed\'? I upgraded BOINC several times to see if something is wrong on that end. I even tested BOINC v6.6.38 meanwhile...

Any more ideas - before wiping CPDN directory?
Plus I am not sure any more that this would help.-

Cheers!
8) Questions and Answers : Windows : Lots of model errors... (Message 38951)
Posted 21 Feb 2010 by Profile cooper
Post:

Hi Les,

of course, av updates are a change. I am aware of the possibility that avast could block something. But not without warning, or moving it to the quarantine without any traces in the logs.

I switched to different models, noticed that some are not available. But that did not help either (there was a hint in some of the sticky messages). Now \'none\' selected, since this is not the reason.

Changed some things, let\'s see...

TNX.
9) Questions and Answers : Windows : Lots of model errors... (Message 38946)
Posted 21 Feb 2010 by Profile cooper
Post:
stderr out message is
Could not launch model process. Last Error=193

which indicates an invalid application. Check that your antivirus hasn\'t quarantined any of the CPDN programs (many users have found the new Norton Sonar scanner to be particularly aggressive).


Hi Thyme,

nothing in the LOGs of my antivirus (avast). Nothing in it\'s quarantine folder related to BOINC/CPDN. I\'d never install N*rton on my personal PC...

Should I try and empty my project folder? The models seem to crash while initializing, within first 2 minutes.

BR,
cooper.
10) Questions and Answers : Windows : Lots of model errors... (Message 38945)
Posted 21 Feb 2010 by Profile cooper
Post:
The errors began 28 January. Did anything change on your PC at that time?

What antivirus, antispyware, and firewall are running on your PC?


Hi geophi,

thanks for your reply. Nothing changed: antivirus as usual (the same more than 2 years), no antispyware, XP firewall.

Models quitting with error began earlier. The first on Aug. 17 (WU 6535960, which crashed on all clients).

Other projects run well, so I can exclude my PC (not O/C anyway).

Unfortunately, in BOINC there is no LOG for each project, just one. And that is not very detailed and obviously rolled over to *.old, until size is 2MB. I did not change config files yet (to increase verbose level).

BTW, all models crash within the first 2 minutes, tried it again today. Also switching to other CPDN model-types did not show success.

What else could I try?

11) Questions and Answers : Windows : Lots of model errors... (Message 38942)
Posted 21 Feb 2010 by Profile cooper
Post:
Hi,

all my workunits fail since some time. When I look at other computers, it is obvious that I\'m not the only one. Some report the same. For example this:
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=6737097
or this:
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=6665065
or this:
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=6664179

BOINC is v6.10.32, but it is the same with v6.10.17. Detaching did not help.

Currently I suspended CPDN, because my downlink is not that fast. And downloading so many WU\'s for nothing is quite a waste of resources. Reaching quota anyway.

I\'d love to continue crunching - but I can\'t.

What could be the problem? I\'m quite sure I did not change anything. BTW, computer # is 985599.

TNX for ideas. Cheers.




©2024 climateprediction.net