Re: Weird network behaviour - can anyone explain it?

[Jacob Anawalt <jacob@cachevalley.com>]

Oliver Elphick wrote:

> We have a machine whose network configuration is in some way wrong, but
> I don't know how.
>
> When it boots, the network is configured correctly, according to
> ifconfig, but it takes forever for things (a deliberately vague word) to
> be processed.  Then it seems to handle a number of requests all at once
> and goes back to sleep for a while.  The effect is illustrated by this
> ping across the local ethernet (other machines on the same net have no
> problems):
>
> #  ping braydb
> PING braydb.somedomain.com (192.168.1.18): 56 data bytes
> 64 bytes from 192.168.1.18: icmp_seq=33 ttl=64 time=0.6 ms
> 64 bytes from 192.168.1.18: icmp_seq=34 ttl=64 time=0.8 ms
> 64 bytes from 192.168.1.18: icmp_seq=35 ttl=64 time=1.4 ms
> 64 bytes from 192.168.1.18: icmp_seq=18 ttl=64 time=17002.5 ms
> 64 bytes from 192.168.1.18: icmp_seq=19 ttl=64 time=16003.6 ms
> 64 bytes from 192.168.1.18: icmp_seq=20 ttl=64 time=15004.2 ms
> 64 bytes from 192.168.1.18: icmp_seq=21 ttl=64 time=14004.8 ms
> 64 bytes from 192.168.1.18: icmp_seq=22 ttl=64 time=13005.4 ms
> 64 bytes from 192.168.1.18: icmp_seq=23 ttl=64 time=12005.9 ms
> 64 bytes from 192.168.1.18: icmp_seq=24 ttl=64 time=11006.5 ms
> 64 bytes from 192.168.1.18: icmp_seq=25 ttl=64 time=10007.0 ms
> 64 bytes from 192.168.1.18: icmp_seq=26 ttl=64 time=9007.6 ms
> 64 bytes from 192.168.1.18: icmp_seq=27 ttl=64 time=8008.1 ms
> 64 bytes from 192.168.1.18: icmp_seq=28 ttl=64 time=7008.6 ms
> 64 bytes from 192.168.1.18: icmp_seq=29 ttl=64 time=6009.1 ms
> 64 bytes from 192.168.1.18: icmp_seq=30 ttl=64 time=5009.7 ms
> 64 bytes from 192.168.1.18: icmp_seq=31 ttl=64 time=4010.3 ms
> 64 bytes from 192.168.1.18: icmp_seq=32 ttl=64 time=3010.9 ms 
>
> --- braydb.somedomain.com ping statistics ---
> 49 packets transmitted, 18 packets received, 63% packet loss
> round-trip min/avg/max = 0.6/8339.2/17002.5 ms
>
> After some considerable time, this effect stops and normal response
> times resume.  (I hope that this will also be the case on this occasion;
> the machine has been running for 5 hours so far.)
>
> Kernel is 2.4.20 SMP, built for this machine.
> I can't identify the network card until the machine starts to respond
> correctly (I am not on site).
>
> The problem began a couple of months back; I do not know of any relevant
> software change.  Since then, the machine has not been rebooted again
> until today.
>
>  
If the other machines on the network were pinging the same remote 
machine just fine at the same time the above is happening then I'd 
suspect: network cables to this computer; port on the hub/switch that 
this computer plugs into; driver for this computer's ethernet card; this 
computers ethernet card. Trying a different port is a quick test, so is 
trying a different ethernet cable.

If the problem is one of the above guesses, you should get the same 
results pinging another local machine while the problem is showing up 
pinging that remote system.

The lag looks exactly like what I have seen in the past on oversold 
service or dslam issues on DSL (routing or network congestion issues).

Good luck.
--
Jacob