At 17:38 19/09/02 -0400, Alan Brown wrote:
>On Fri, 20 Sep 2002, Simon Byrnand wrote:
>
>> The only way a stale .pop file would be left behind indefinately is if the
>> popper process itself crashed, or was forcibly (-9) killed, or the
>> operating system crashed.
>
>Or Qpopper flags are set to leave them behind. They're another way of
>seeing whan a user last popped.

Would that prevent new sessions from logging on later though ? And if so,
what would be the point ?

It sounds like in his case the stale .pop files are preventing him from
logging in again, so I doubt thats the case here.

It needs to be established whether they're really stale, or whether the
popper process is just taking a long time to exit after a session is lost.

I found that it could take anywhere from 5 to 15 minutes for a session to
time out if the user "dissapeared" completely (eg modem hangup)

In my case (Linux 2.2.21) I found the following settings eased the problem
a great deal:

echo 3 >/proc/sys/net/ipv4/tcp_retries1
echo 5 >/proc/sys/net/ipv4/tcp_retries2

The default settings were 7 and 15 respectively. Now a disapearing client
causes qpopper to timeout usally between about 2 and 3 minutes. (And
doesn't seem to negatively impact other programs.... since it is a global
tcp setting...)

Regards,
Simon


Reply via email to