Randomly disappearing server

[DusteD]

This problem is really getting to me.

The server
Debian Stable
NIS/NFS/Samba/Apache

The Problem
The server suddenly and randomly disappears, like, its just puff, gone!
Cant log in over ssh, nfs clients freezes, it dosent respond to ping.
Then sometimes up to 10 minutes after, pop, there it is, it starts responding to ping
and people can log in again..

I havent been able to provoke it to disappear (by stressing nfs, loggin in and trying to bruteforce nis accounts :P)

Theres NOTHING in the logs, or on the terminal, or in dmesg.
ifconfig shows no lost packages, even after a "freeze"

when i am in front of a client and it freezes, if i run into the server room and log in as root, theres no messages, but i cant ping out from the server either but ifconfig shows it has an ip address..

I have tried moving the server to another room with better networking, ive tried changing the nic (3 times!), ive tried diffrent cables, im totally beat here...

Please, if you find it in your heart to come with ANY advice, ill be very happy

Thank you.

[antidrugue]

but ifconfig shows it has an ip address..

Is it always the same?

[DusteD]

Hi, thank you for your answer, yes it is the same

i encountered (once) this error message:
RPC: bad TCP reclen 0x39366436 (non-terminal)
I lil googling tells me its something to do with nfs and/or tcp
i wouldnt expect this kind of behavior from a sarge install.
The clients are testing, but still, you shouldnt be able to crash
a server just by talking wrongly to it..

[antidrugue]

I lil googling tells me its something to do with nfs and/or tcp
i wouldnt expect this kind of behavior from a sarge install.

I remember having some networking issue with the default 2.6 Sarge kernel. I fix it by compiling my own kernel (I used 2.6.12.x, the last one that works well without udev).

[DThor]

Can you repro without running NFS? - maybe you're not in a situation where you can practically take that offline, though. I'm hearing about NFS issues too, especially with large file transfers - when you changed the NIC - was it replaced with the same driver-type? I'm reading about issues with aic7xx drivers in the kernel around that version.

Certainly a kernel upgrade would be the next thing I'd investigate. Sure sounds like a kernel/driver issue, anyway.

DT