linux clients crashing

[Scoofy12]

I had several (9/24) of my linux clients crash today, giving this error message. any idea what it means?

latest client, RH linux 7.2 on a Dell poweredge (7100 i think) dual p3/800, with one client running per box. client runs quiet and no progress updates in a screen.


========================[ Apr 3, 2002 6:38 PM ]========================
!!! Contact between residues LYS and GLN too close !!!
FATAL ERROR: [065.002] {asnio.c, line 1096} 5efroq0c_protein_0020009.val.bz2Input

tag mismatch on Ncbi-mime-asn1
FATAL ERROR: [065.002] {asnio.c, line 1096} 5efroq0c_protein_0020191.val.bz2Input

tag mismatch on Ncbi-mime-asn1
FATAL ERROR: [065.002] {asnio.c, line 1096} 5efroq0c_protein_0020048.val.bz2Input

tag mismatch on Ncbi-mime-asn1

[Brian the Fist]

For one thing that is an output file not an input file so why is it trying to read it in the first place? Have you modified the program/scripts in any way?
At exactly what point in the program does it crash with that message?

[Scoofy12]

sorry, let me clarify. the clients crashed (or at least terminated abnormally) at various times overnight, and what i pasted is, as you know, out of error.log. presumably this one went down at 6:38 PM and wrote to error.log before it did so. i'm generally unable to see the stdout of the program after it crashes.
here's the setup:
all the client install files for each system are hardlinked together from a master set, but each client has its own working directory. (ie in my home directory ~ there exists ~/masterproj and then each computer has access to my home directory (via NFS) and runs a client in ~/projX, where x is the computer number. all the installed client files in ~/projX are links to those in ~/masterproj, but each client runs in its own directory, and creates the lock, log and bz2 files in ~/projX)
rather than using the foldit script, each client runs the following shell script in a program called screen (which allows you to start a tty and then detach and logout without disturbing it)

#/bin/bash
#echo Starting Distributed Folding in proj$1
cd proj$1
./foldtrajlite -f protein -n native -qt -g 0
cd

this may seem convoluted, but it saves on disk space and assures that each client gets its own working directory.

are you saying that 5efroq0c_protein_0020048.val.bz2Input
is an output file? i don't know why its trying to read it. i've restarted the clients and deleted the leftover bz2 and log files so we'll see if they crash again. thanks for all your work :)

[Brian the Fist]

I have mentioned this before but Ill say it again. Please do NOT make links, hard or soft, to run multiple copies of the software. You must install N copies to run it N times. I don't want to sound harsh, and it is certainly nothing personal against you, but if you go ahead and do this, please don't then tell me it doesn't work.

Try running a single copy of the client in a single directory on a local machine, and if you still have problems Ill be glad to help, but I cannot support what you are trying to do there, sorry.

[Scoofy12]

hehe... no problem, i totally understand.
i did that and it seems to be working ok.
thanks!