At 2:18 PM -0600 11/14/00, Kip Turk wrote:
> We running about 1700 users on a PII 500 with 512M RAM running RH 6.0 (I
> think, been a while since I was at the console). Lately, it has started
> spooling out of control during peak hours. Last night, it had 325 popper
> requests (we normally max at 20) and 75 zombies. I've upgraded the system
> to 3.1, but it hasn't fixed the problem. These are the options I
> configured with:
>
> ./configure --enable-specialauth --enable-temp-drop-dir=/var/spool/poptemp
> --enable-debugging --enable-servermode --enable-shy
> --with-log-facility=LOG_LOCAL0
> --enable-timing
>
> I've started to see these types of messages in the local0log:
> #tail -f local0log | egrep -B 1 '\<clean=[1-9]\>'
> Nov 14 14:17:58 brazos popper[17409]: Stats: osg 0 0 509 4448005
> 204.94.234.10 204.94.234.10 [pop_updt.c:228]
>
> Nov 14 14:17:59 brazos popper[17409]: (v3.1) Timing for [EMAIL PROTECTED]
> (error) auth=0 init=1 clean=1 [popper.c:289]
> --
> Nov 14 14:18:08 brazos popper[17419]: (v3.1) Timing for
> [EMAIL PROTECTED] (normal) auth=0 init=1 clean=0 [popper.c:289]
>
> Nov 14 14:18:10 brazos popper[17412]: (v3.1) Timing for [EMAIL PROTECTED]
> (normal) auth=0 init=1 clean=3 [popper.c:289]
>
>
> Can anyone see anything I've misconfigured? Can you explain why one pop
> gives an error and the other shows normal? Can you recommend anything to
> fix this problem? It may not be a qpopper issue, but it seems to be the
> guilty party as it has the highest number of process requests and zombies
> when the system starts to die.
>
> Thanks for the assist,
The "(error)" versus "(normal)" refers to how the session ended.
If the connections aborts, or the client just shuts down the
connection without sending "quit", Qpopper reports this as an error.
If you're seeing zombie processes when Qpopper is run from inetd
(as opposed to standalone mode) then I suspect a problem with
inetd. I'm not aware of anything Qpopper can do to cause or avoid
zombie processes when it is forked by inetd.
