Pentium M, hmmmmm. Are these laptops ? Could you have heat problems if so ? Is it the same machines consistenly ? Do they quit at random times or is there a pattern ? Just some thoughts.

Cheers

There are no heat problems cpu's run between 40 and 60 degrees Celcius, spec is above 100 degrees celcius. These systems are industrial pc's which should produce not to much heat. To check if the pc's are ok i stress them with boinc. The only way i'm sure processor is running full speed and there is a lot of memory access.

pattern, no, some systems it takes 5 day's before the first problem, others at the first day. It is more or less random. But only on these machines at the moment. the Celeron running windows have no problems, and the AMD machines have also no problem.

Project switches are normal on 1 hour bases. To be sure it is not the project switching thats wrong, i took the systems off SETI, SIMAP has no work so there is nothing coming from there at the moment. So its only WCG that is crunching.

linux version is LTSP - 4.2, which is installed on a Suse 10.0 server.

log shows nothing special, i'll check if i can get on remotely.

top - 23:42:33 up 12 days, 9:06, 0 users, load average: 0.00, 0.00, 0.00
Tasks: 40 total, 2 running, 38 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.3% user, 0.0% system, 0.0% nice, 99.7% idle, 0.0% IO-wait
Mem: 246988k total, 136972k used, 110016k free, 0k buffers
Swap: 0k total, 0k used, 0k free, 45648k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1 root 15 0 1428 476 420 S 0.0 0.2 0:00.71 init
2 root 34 19 0 0 0 S 0.0 0.0 0:00.00 ksoftirqd/0

-bash-2.05b# uname -a
Linux ws203 2.6.20.9-ltsp-1 #1 PREEMPT Wed May 9 06:08:26 EDT 2007 i686 unknown

-bash-2.05b# ps ax | grep wcg
21505 root 294508 S wcg_faah_autodock_5.42_i686-pc-linux-gnu -d
21506 root 294508 S wcg_faah_autodock_5.42_i686-pc-linux-gnu -d
21507 root 294508 S wcg_faah_autodock_5.42_i686-pc-linux-gnu -d
21508 root 294508 S wcg_faah_autodock_5.42_i686-pc-linux-gnu -d
21509 root 294508 S wcg_faah_autodock_5.42_i686-pc-linux-gnu -d
22891 root 1560 S grep wcg


from messages tab of boinc manager:
Tue 23 Oct 2007 06:05:52 PM CEST|World Community Grid|Task faah2534_ZINC01669772_xmd03400_01_0 exited with zero status but no 'finished' file
Tue 23 Oct 2007 06:05:52 PM CEST|World Community Grid|If this happens repeatedly you may need to reset the project.
Tue 23 Oct 2007 06:06:01 PM CEST|World Community Grid|Restarting task faah2534_ZINC01669772_xmd03400_01_0 using faah version 542

tuxnl, we need the message log to be able to diagnose.

How does this "preempted" manifest itself? If a job is not finished and another is started e.g. due project switching it would show that. Reads like not only WCG is on those clients, thus whatever switching time, 60 minutes is default, it would do that. Recommended is to run switching at e.g. 4-5 hours, so lots of checkpoints get saved. BOINC will ensure that each project gets it weighted share of time.

The efficiency 0.000% is presumably what you see in the Tasks window Progress column. Are those tasks counting up CPU time and using any. The sciences run at 'nice' priority and understand in Linux some setting has to be changed to get it to go at really using the full idle cycles.... cant remember momentarily what that setting was.

BUT, the Why only WCG is puzzling.

The one message you posted earlier related to "zero status but no finished file" is covered in an FAQ. One cause is the system/OS automatic time synchronisation. BOINC does not like times adjusted while running, particularly backward, even a second. I've switched it off and manually adjust time if it gets too much off from internet time.

he Sekerob also here.

The preempted, could be SIMAP, they dumpt some work on these systems, i'm closing the work fetch for SIMAP now. That explains it

Task switch is 60min. write to disk is 60 min, to reduce network stress and make sure this could not be the problem. If at all, use switch times of 4 hours plus on multi-project machines.... it's really more efficient in several ways. The Disk save is 60 seconds, not minutes. I use 600 seconds, again for increasing efficiency

CPU efficiency 0.000% is from boincview. After my restart it goes back up to 100.0 or 99.96.Yellow or Red lines then, BV is my one stop shop to watch the lot ;D

i'll look in to the time sync, don't mind what the time is on this cluster. Let us know it if makes them messages disappear

Which log do you want, in the message tab of the boinc manager there is no logging ahich shows this problem.Those showing in the message tab and ideally 5-10 from before to 5-10 after the 'suspicious' items or times hanging.... still the question... is it HPF2 baulking or other WCG projects too?

Just reading about Ubuntu Gutsy (7.10?) and the much better font rendering in the BOINC manager. Going to get one of those LiveCD's burned and see how that takes liking to me. After Vista exposure now, the absence of WoW is deafening.