Topic: e-smith server stops responding

[Adrian]

E-smith stops responding (ie no email or web access). This can happen within 2 to 3 hours of a reboot or 2 to 3 days after a reboot. After a reboot the machine is fine. I have searched the logs and there doesn't appear to be anything to indicate a problem. 'top' indicates CPU 98% idle and 87 Mb free memory. Incoming emails are still being accepted. I am using SOCKS to access a remote socks server, and also mounting drives from a remote NFS server. The machine is a dual homed (intel pro 10/100 NIC's) PII 233 128Mb, access to the internet is via a router to an ISDN line.

Any help on this matter would be appreciated

Adrian

[John]

I am getting the same problem on 2 servers. both using a router to DSL...   one started this after being upgraded to ver. 4.1.2.  the other is running 4.1.1

[Christopher Smith]

Not sure if this is related to your problem or not, but I experienced the same thing with my machine a while back.

Turns out (according to Shaw tech support) that there's a bug in Red Hat that occasionally interferes with the DHCP client if you are currently logged in (either directly on the terminal or through telnet) as root. It does not seem to be affected by logging in as a user and then su to root.

I have not been able to verify this bug, but the 2 times it happened to me I was logged in as root. I stopped doing that, and it hasn't happened since.

What seems to happen is the DHCP client cannot speak to the DHCP server, so it assumes it can keep the IP address. However, if your DHCP lease expires on the server, there's the risk of it being given to someone else. Now both your machine and the new machine think they have the same IP (but you're outta luck).

Solution: give a user shell access, log in as that user, su to root, bring down ethX and bring it back up.

Hope this helps.

[Adrian]

Thanks for the advice Christopher. The e-smith server has a static IP.

All the desktop machines have static IP's. We do have a Win NT DHCP server which has a limited number of addresess to allocate, which are used by our notebook users.

I have turned off Squid and Httpd-e-smith to see if that makes a difference, although turning off httpd-e-smith also means we lose Webmail.

Adrian

[Charlie Brady]

Christopher Smith wrote:

> Turns out (according to Shaw tech support) that there's a bug
> in Red Hat that occasionally interferes with the DHCP client
> if you are currently logged in (either directly on the
> terminal or through telnet) as root.

We do not use the same DHCP client as RedHat, so this is not likely to be the problem.

> Solution: give a user shell access, log in as that user, su
> to root, bring down ethX and bring it back up.

If that helps, then it could indicate a network driver bug, and/or buggy NIC hardware. Replacing the NIC with a different brand should make the problem go away if that is the case.

Regards

Charlie

[Darrell May]

Have you tried a new hub/switch?

[Ron LeVine]

I recently had this issue as well with the notable exception of the whole server appearing to go to sleep. I wound up being a bad motherboard. The E-Smith is very solid and when the server does something wonky, I look to hardware issues.

Ron