Runaway zombie processes after init upgrade (anyone seen this before?)
On Friday I installed some updates on 4 Sarge boxes and 1 Woody box. Among
the upgraded packages for each server was "sysvinit", the package that
contains /sbin/init, among others. On each machine I noticed the an entry
in the logs that said something like "trying to re-exec init" (which I've
learned is apparently necessary after init is upgraded).
On two of these machines right at the point init was re-exec'ed I started
getting zombie processes (according to a cron job I have setup to log the
output of ps every 5 minutes to a file). One of the machines was the Woody
box, and the other was a Sarge box. Apparently the number of zombie
processes (which happened to be sendmail processes) kept increasing until
finally the two systems could not fork any new processes and I had to power
cycle them (today...it took 48 hours for them to become unusable).
Strangely enough the other 3 Sarge boxes I have are all running Sendmail as
well, and they all had init upgraded and re-exec'ed, but they are all
running fine. I did some googling and discovered a couple of other people
saying that this has happened to them (init starts leaving zombies after an
upgrade) and the only solution is to power cycle since you can't even shut
the machine down gracefully when this happens.
I was just curious if anyone has seen this before...is this just something
that happens from time to time, or would you consider this a bug in the
upgrade process...? I'm not sure if I should bother to report it, because I
can't find anyone else who's had this problem recently and it looks like
it'd be one of those things that would be a nightmare to reproduce...
I'd appreciate any feedback...