On Thu, 23 Mar 2000, Aviram Jenik wrote:
> A friend of mine
yeah, it's always "a friend" ;) (just kidding).
> has been having strange server crashes (about once a week
> in the last few weeks). His server was extremely stable until lately, and it
> doesn't do anything special except POP3 (qpopper), SMTP (qmail), and web
> (apache).
no linux system every does "a special thing". they simply run and work.
on the other hand - machines crashs sometimes, and that's while not doing
"special" things - they just have a regular problem.
> My immediate suspect was qpopper, which he runs through inetd and tcp
> wrappers (according to him, the last thing on the log before each crash is a
> POP3 request).
why would qpopper cause the machine to crash? this is hardly ever the
case. why did you suspect qpopper anyway? cause it had the last entry in
the log? that's quite irellevant. the fact that it managed to write the
log entry is actually evidence that it's probably something else.
> I wasn't sure whether to tell him that the problem is
> qpopper, inetd, or maybe a freak hardware problem.
according to the details of your problem, and past experience (my own, and
other people on the list and off it), this is most likely a hardware
problem. check out the following:
1. the CPU fan stopped working. this will cause the CPU to heat up and
eventually hang.
2. the CPU FAN working too slow, due to too weak power supply, for the
hardware in use. this is possible if some new ahrdware was added
eventually, that consumes more power. check this out anyway.
3. memory getting bad. it may cause the maching to hang when it tries to
access the faulty memory. not always will it acuse a kernel panic
(btw, you didn't exactly say if you see a kernel panic, or a coplete
hang. not that you always can see that - if the server is in X mode, a
kernel panic (written to the first vritual console) won't be visible.
4. motherboard problems can cause sch a thing as well.
> My only recommendation to him was to change inetd to xinetd, but I wanted to
> hear other opinions since I'm almost clueless on this.
one question - is there any evidence in the log files for too much load on
the amchine, before the crash? cause that (and kernel bugs) are the only
thing that might somehow cause a system hang (and even with large load,
there are ways to notice this, e.g. even on large loads, the machine still
answers 'ping' messages sent from other machines - is that the case?).
guy
"For world domination - press 1,
or dial 0, and please hold, for the creator." -- nob o. dy
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
