NNTP periodic failures on ISC UNIX

[Heiko Schlichting]

paul at actrix.gen.nz (Paul Gillingwater) writes:

>We have a problem with our NNTP, and would appreciate any helpful
>suggestions.

We have a similar configuration and the same problems.

>Here's our configuration:
>
>386/33MHz with ISC 2.2.1, running TCP/IP 1.2 on a WD 1003E LAN card.
>We're connected via a PCRouter (software running on another PC) to a
>local University Internet node, where we receive mail and news via
>SMTP and NNTP.  We have 10 Mb of RAM.
>
>We're using NNTP version 1.5.11 (with patches for ISC UNIX).

We use a 386/33 MHz with WD 8003E and ISC 2.2 + NNTP 1.5.10.

>The problem is that two or three times a day, the NNTP appears to
>lock up.  This also causes mail to stop, so it seems to be a problem
>that restricts a resource used by both SMTP and NNTP (perhaps
>streams buffers?)

We have this problem too. 
Do a "telnet 127.0.0.1 119" only gives a "trying..." and nothing happens.
There are similar problems with SMTP ("telnet 127.0.0.1 25") some times
but it seems to be independant from NNTP problems. At one time NNTP locks 
up and at another SMTP locks.

>We've tried shutting down the LAN, then restarting it by changing the
>init level, but this does not seem to work.  We know that the
>PCRouter is fine, because we can still TELNET or FTP over the LAN
>(sometimes).  We can usually PING the PCRouter successfully too.
>This has meant that the only way we can restart the connection is to
>do a complete powerdown and reboot.

If NNTP locks up we have to shutdown too but for the SMTP problems there
is a workaround:

kill all running "sendmail -bd" processes and restart the sendmail daemon
with "/usr/lib/sendmail -bd -q10m". This works in most cases but you 
get a bind() error sometimes. In this case only a complete reboot helps.
We use smail 3.1.20 - I don't know if this works with the ISC sendmail too.

I think the problem is that there are a lot of TCP/IP process is the
status CLOSED and they didn't terminate. 
This is *NOT* a special problem of NNTP. It is a general TCP/IP problem
with ISC.

Look at our netstat output - there are CLOSED processes by uucp (over TCP/IP),
telnet, smtp and ftp too:

Active Internet connections
Proto Recv-Q Send-Q  Local Address          Foreign Address        (state)
tcp        0      8  methan.chemie.fu.1660  ki1.chemie.fu-be.uucp  CLOSED
tcp        0      0  methan.chemie.fu.nntp  crane.aa.ox.com.4337   LAST_ACK
tcp        0      0  methan.chemie.fu.1484  athene.uni-pader.ftp-d LAST_ACK
tcp        0      3  methan.chemie.fu.1430  taurus.tat.physi.4242  CLOSED
tcp        0      1  methan.chemie.fu.1416  taurus.tat.physi.telne CLOSED
tcp        0      8  methan.chemie.fu.1413  ki1.chemie.fu-be.uucp  CLOSED
tcp        0      0  methan.chemie.fu.telne gibb.math.fu-ber.2879  ESTABLISHED
tcp        0      8  methan.chemie.fu.1215  ki1.chemie.fu-be.uucp  CLOSED
tcp        0      8  methan.chemie.fu.1162  ki1.chemie.fu-be.uucp  CLOSED
tcp        0   1846  methan.chemie.fu.telne troll.cs.tu-berl.2778  CLOSED

And after a while - this will lock TCP/IP and makes a reboot
necessary.

Bye, heiko.
-- 
 |~|    Heiko Schlichting                   | Freie Universitaet Berlin 
 / \    heiko at fub.uucp                      | Institut fuer Organische Chemie
/FUB\   heiko at methan.chemie.fu-berlin.de    | Takustrasse 3
`---'   phone +49 30 838-2677; fax ...-5163 | D-1000 Berlin 33  Germany