ISP connection slowdown after a couple of days

ISP connection slowdown after a couple of days

Post by Michael Klei » Thu, 22 Jul 1999 04:00:00


I'm running a small LAN with two dual boot Linux/Win0.95 and a Linux
gateway running kernel 2.2.9 and ipchains 1.3.8 for firewalling and
masquerading (no proxies). Dial-on-demand connections via ISDN are made
with ipppd 2.2pl10.

After a couple of days there is a noticable slowdown with certain
services, especially DNS, but also POP3, NNTP and telnet. I don't know,
if it's caused by the gateway's uptime or by the number of connections
made, though. After reboot, everything is fast again for a while...
I guess, this slowdown is caused by high latency times, not lack of
bandwidth (hope you understand what i mean, sorry for any bad english).

With 2.0.36 and ipfwadm, there were no such problems, but I used ipfwadm
only for masquerading but no firewalling at all (gotten paranoid:) ).

Furthermore, I'd like to run a caching-only nameserver on my gateway
(ISP's nameserver as forwarder in /etc/named.boot, nothing else). However,
forwarded DNS querys seem to be sent over UDP, not TCP, which bites
ip_dynaddr, is this correct? Or, in other words, in which cases is
either UDP/TCP used for DNS?

Any solutions/hints appreciated, TIA

(please Cc any follow-ups, I'm not a frequent reader)


1. Slack slows down after a couple of days?

slack 7.1
2.2.16 kernel
486 12M RAM
3 NICs
NO X-server
Used as gateway/router/masq/DNS server box

About 2 weeks ago pkgtool did me in and I had to do a complete new install.
I took advantage of the situation and re-formatted and re-installed
everything so I know I'm working with a clean install.

The basic problem is that after several days I notice the box slows down
almost to a grinding halt.  I'll go to login using telnet and it will take a
minute or so before "Login:" pops up.  After re-booting everything is fine.

Before I did the re-install I never had to touch the box.  It ran for weeks
on end and I never had any problems, didn't even look at it.  Now it can't
go for a couple of days before it's dragging.

I cut/pasted a copy of the screen while top was running after having it up
for a day and then I did it again when it was dragging and there isn't any
big difference.  It is using a little more memory, but the CPU usage is
still down around 93% idle, the same as when it was first started.  I'll be
happy to send anyone that wants a copy of the 2 printouts of top.

I begin to get complaints that people can't send me email which is routed
through my slack box to my email server, I suspect that it just gets so slow
that the email server(s) time out.

Any ideas and/or suggestions?

2. installing SO50 in KDE (for dummies :o( )

3. dhcp not starting since a couple of days

4. Can't find my ISDN card

5. ISP connection: starting PPPD on ISP's machine

6. kylix (c/delphi) under RTlinux?

7. pppd demand dialing mode work great, for a couple of days then it stops.

8. email Autoreply on HPUX 10.20??

9. ISP connection sharing via a PPP connection with NULL-MODEM

10. ASCII connection not allowed to my ISP with pppd connection

11. PPP connection slowdown?

12. MASQ causing connection slowdowns?

13. Informix problems afters Y2K Solaris 2.5 patch