Jody L. Baze said:
This is a well-known problem with 2.6. I've had a bug report
filed with Sun since, oh... November. At least as far as I know, there is no
patch available. It really has nothing to do with rotations, per se, it has
to do with HUP'ing syslogd several times in a row. A race-condition seems to
exist in the code which allows a rather substantial period of time to go by
without a signal handler for HUP. The result, as you've noticed, is that the
default HUP action is taken; syslogd terminates.
What I did:
I replaced the syslogd on my 2.6 boxes with an older syslogd from a 2.5.1
machine. It no longer dies when I do all the kills as part of rotating
logs. My 2.5.1 syslogd had broken when I installed a security patch from
Sun so I had gone back to the syslogd from before that.
-- Chris Marble - HMC UNIX Systems Manager, chris_marble@hmc.edu My opinions are my own and probably don't represent anything anyway.