Boinc/Einstein crashing

norobro · 10-10-2006, 09:20 PM

Hello all.

I’ve been running Einstein/Boinc on Debian Sid for about a year and a half. Recently my work units have been crashing with the following error:

Code:

_SSEgas.c
2006-10-07 22:03:57.7242 [CRITICAL]: At lowest level status code = 0, description: NO LAL ERROR REGISTERED

2006-10-07 22:03:57.7243 [CRITICAL]: APP DEBUG: Application caught signal 6.

2006-10-07 22:03:57.7243 [CRITICAL]: Stack trace of LAL functions in worker thread:
2006-10-07 22:03:57.7243 [CRITICAL]: TestLALDemod at line 80 of file /home/bema/einsteinathome/CFS/EaH_build_release_einstein_S5R1_4.17/extra_sources/lalapps-CVS/src/pulsar/FDS_isolated/CFSLALDemod_SSEgas.c
2006-10-07 22:03:57.7243 [CRITICAL]: At lowest level status code = 0, description: NO LAL ERROR REGISTERED

Einstein Site: Results

I checked the logs and there are no entries at the time that the program crashed.

Also, I posted to the Einstein problem and bug reports page and got the reply that “something on your machine keeps trying to terminate the app.” It’s trying and succeeding!

My system:
Processor: AMD Athlon XP 2700+
Memory: 512 Mb
Debian Sid
Kernel: 2.6.18 self compiled.

Thought I’d request help from the experts. Any ideas?

TIA for any help.

Norm

Dutch Master · 10-10-2006, 09:42 PM

Well, the Einstein folks gave you a hint: something is killing the app! Next job: figure out what makes the app to stop, i.e. what programs does it depend on that quit working. (like a chained dependency: app x depends on prog y to run, but process w kills prog y, causing app x to shutdown) Figuring that one out is a PITA (pain in the *ss

) but probably the only way to get to the bottom of your problem.

norobro · 10-11-2006, 09:00 AM

Dutch,

Any suggestions on how to go about troubleshooting this?

Don't know how familiar you are with Einstein. I'm using a downloaded binary Boinc and it automatically downloads the Einstein app if there is a new version. It (Einstein app) is not open source.

With the absence of log entries, should I start killing programs? The last three WUs have run for four hours before terminating. That makes for a lot of wasted cpu cycles.

I haven't performed a dist upgrade in about a month, so I'll start with that.

The puzzle to me is that it just started in the last week. The last thing that I have done is compile the 2.6.18 kernel on Sept. 20. If the problem is there, it took over two weeks to show up.

Thanks for the reply!

Norm

Dutch Master · 10-11-2006, 09:40 AM

As you already anticipated, I'm not familiar with either Boinc or Einstein. Finding the culprit requires a methodical approach. First, figure out what the dependencies of Einstein and Boinc are. Was the binary a .deb package, 'cause if it was the dependencies should be in the package description. Next, check wether the installed versions of those dependent packages are suitable for Boinc/Einstein. As you've said that Boinc d/l's a new Einstein package automatically, I suspect the app expects a certain package installed but you still have an older, not quite compatible version. However, your remark on the uptime on those WU's I don't exclude the possibility there might be hardware issues (i.e. upcoming failures) causing these shutdown's.