SUMMARY: Solaris 10 user processes

From: Lisa Weihl <lweihl_at_cs.bgsu.edu>
Date: Wed Sep 27 2006 - 11:01:35 EDT
Apparently I've missed a system setting for /etc/system that's been 
around as long as Solaris.

Restrict the user processes with the following line in /etc/system:

set maxuprc = <num>

I don't know how I've managed to not get hit by fork bombs 
before!  Although most of the time students are using machines at the 
University level and not our local dept. machines.

As to whether the machine crashed or was just unreponsive.  I guess 
it was just unresponsive but it was to the point that even the window 
manager was dead (yes, I know it's not normal to have a graphics card 
in a server:-) and the screen was just white.  I had to power cycle 
from the ALOM to reboot.

Thanks to the following for all their helpful advice!
Casper Dik, Rich Teer, Ric Anderson, Rich Kulawiec, Kevin Prigge, 
Eric Grejda, Dana Hudes, Helmut Kreft, Deb Crocker, Martin 
Presslaber, Damian Wiest, Peter Gray and Francisco

Lisa

***************************** ORIGINAL MESSAGE 
***********************************
Hi all,

Very new to Solaris 10 and it's new way of handling a lot of 
things.  I have a new Sun Fire T2000 running Solaris 10 6/06.  I 
recently setup accounts for an advanced networking course on this 
server.  The students are writing programs that are using fork and I 
believe some of them are causing infinite loops.  Twice in the last 
two days the server has crashed.  A power cycle brings it back up 
each time.  There are only two helpful messages in the log:

On the 24th I had:

/etc/svc/volatile:File system full, swap space limit exceeded

and also lots and lots of the following....

sendmail [306]: rejecting connections on daemon NoMTA4:load average: somenumber

where somenumber ranged from the hundreds to the thousands.

Last night, the 25th the server crashed again but I only had the 
sendmail error messages listed this time.  I figured the "file system 
full" messge from the 24th just had to do with someone managing to 
fill up /tmp.

This morning the faculty member for the course wrote a program to 
fork a lot of processes, ran it and immediately the machine hung 
requiring a reboot and the familiar sendmail message was in the log.

Clearly these crashes are tied to users having runaway fork calls but 
I've never see a server allow a user to fork enough processes to 
bring a machine down.  Can anyone steer me in the right direction for 
commands to use and things to check to get to the bottom of this problem?

Thanks,
Lisa
**********************************************************************************
Lisa Weihl, System Administrator          E-mail:      lweihl@cs.bgsu.edu
Department of Computer Science          Office:        Hayes 225
Bowling Green State University              Phone:     (419) 372-0116
Bowling Green, Ohio 43403-0214           Fax:         (419) 372-8061
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Wed Sep 27 11:02:14 2006

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:44:01 EST