On Sat, 3 Jul 1999, Doug Lumpkin wrote:
> It's running on a Redhat 6.0 box. It hasn't happened again thankfully, but
> it's not a production server and has a very small load... What would be a
> recommended procedure for diagnosing the exact problem the next time it occurs?
There are quite a number of different things you can do.
* Attach strace to the process, when it hangs. This should tell you if
it's stuck in some system call, or if it's looping. RTFM "man strace".
* If strace doesn't offer much of clue, attach the process to your
favorite debugger. Pray that you can get a strack trace out of it, and
see exactly where it's hanging.
* Take a look around in /proc/pid#, to see what the process is doing. RTFM
"man proc". You'll be able to see what files the stuck process has open,
what signals it has blocked, etc... You can find lots of usefull stuff in
there.
Note: djb's default Makefile strips out all debug stuff out of the
executable, so a debugger won't be much good unless you recompile
everything with debugging turned on.
P.S. No problems with Qmail 1.03 on multiple machines, with
kernel-2.2.5-22
