[ILUG] Thorny networking problem partially solved

[Niall O Broin]

I issued a plaintive cry for help yesterday, and I'm sorry to say was left
walking in the wilderness. Not that there wasn't a number of suggestions,
just that none of them were any use :-)

However, I got 2.4.18 last night and that has cured the sshd weirdness
(sshing into the box killed it dead, leaving keyboard LEDs flashing) and I'm
putting that down to NIC drivers - this box has dual integrated Intel EEpro
100 and this chipset has issues, to say the least of it. Intel have released
a driver which seems more solid (at least in the past) than the stock kernel
one but they've issued it under some arse of alicense which means that it
can't be included in the mainstream kernel. 

However, the NFS mount weirdness remains (an NFS mount from another local
box takes 5 minutes to complete, and an strace shows it hanging in a mount
system call). It really sounds like a DNS issue, particularly because of the
5 minute delay BUT DNS shouldn't be involved - I have 

hosts:  	files dns

in /etc/nsswitch.conf and I have the nfs server's hostname in the client's
hosts file and ping, ftp, ssh all connect immediately. And anyway, it's just
as slow when I use the server's IP address. This actually isn't hugely
important in that I now know that it's not hanging, it's just waiting, and
the box in production won't be using NFS mounts anyway but it's still
bugging the hell out of me.

Just had a thought - added my ISDN box's address as default router to the
troublesome NFS client, added my normal nameservers to its resolv.conf so it
can now happily ping by name the outside world, as can the NFS server -
makes no difference to the problem whatsoever - as I type, an NFS mount
command sits hanging.


Niall