Arch Server Becomes Unavailable

Carlwill · 2010-03-08 13:26:37

This is a weird one for you guys...

I have taken 4 different kinds of operating systems (Linux) and they all have the same problem. The problem begins when I build a server on a Dell PowerEdge 1855 blade. The system builds and I install Apache or Email. Basically once the server is running, people report they can't get to the server via HTTP or what ever service the server is hosting. I then attempt to ping the server...no response via ping. I attempt to SSH into the server...no response. It's as if the server is gone.

I walk into the server room and sure enough the machine is up and running fine. I check the cables and everything looks to be fine. I can login directly at the console and then when I ping the Internet directly from the server, it basically wakes up the NIC or something and then everyone can again access the server. It's as if the NIC goes to sleep. There is no option in the BIOS of a Dell PowerEdge server to shut the NIC down at idle but it happens when the system idles for like 10 minutes or so...

I then reloaded a PowerEdge 2950 (new server) and the same thing happened. How can this happen on two different pieces of hardware. These aren't home made servers either. These are enterprise class machines. I don't get it. The way I have been able to prevent it from happening is when I am logged into the server directly at the console, I start a continuous ping to something and it basically just keeps the NIC awake and active. Super ghetto but it works until I can figure out why this happens.

The only thing these two servers have in common besides the problem is they're both connecting to the same Cisco / Linksys branded SRW 2016.

Anyone ever see anything like this or know what could be causing this problem? Other servers on the same kind of hardware and same switch don't have it. Just a select few.

I don't think it's the NIC or NIC drivers / firmware since the current server uses a Broadcom NIC and the previous one has a Intel NIC. Both had the same exact issue. Could this be the unmanaged switch or something with ACPI?

crankyadmin · 2010-03-08 13:37:36

Pastebin the output of dmesg.

Carlwill · 2010-03-08 13:51:26

crankyadmin wrote:

Pastebin the output of dmesg.

http://pastebin.com/6byDay10

crankyadmin · 2010-03-08 14:01:24

I've seen it before, but you could see the interface coming up & down in the output of dmesg. Which in your case you can't. I would pretty much rule out ACPI putting the interface into sleep, and start looking towards the kernel. Are you using a custom kernel or prebuilt, can you try using a different kernel?

Carlwill · 2010-03-08 14:11:51

Just using the pre-built kernels. Nothing custom at all actually.

Arch Linux

#1 2010-03-08 13:26:37

Arch Server Becomes Unavailable

#2 2010-03-08 13:37:36

Re: Arch Server Becomes Unavailable

#3 2010-03-08 13:51:26

Re: Arch Server Becomes Unavailable

#4 2010-03-08 14:01:24

Re: Arch Server Becomes Unavailable

#5 2010-03-08 14:11:51

Re: Arch Server Becomes Unavailable

Board footer