NICs become unresponsive

[rjs]

We have a 4-server XPA W2K SP3 farm. All are Dell PowerEdge 2550s with embedded BroadCom NetXtreme NICs. We have the servers set to reboot on a nightly basis. Lately (last 2-3 weeks), one of the 4 servers becomes unavailable sometime after the reboot.

When this happens, you cannot ping that server. If you log directly into the Console of that server, you cannot ping out. The NIC lights show connected as does the corresponding Ethernet switch lights. The server NICs and corresponding Ethernet switch ports are specifically set to 100Mbps, half duplex. A subsequent reboot of the server resolves the problem, but am mistified why there is sometimes (very intermittant) the apparent loss of network connectivity. We have no other network issues on the servers.

Any insight or help would be appreciated.

R.Sobelman

http://www.tracklogins.com

 

[KZimmerman]

i assume you have the servers set up to boot in verbose mode? Software is unfortunately not 100% guaranteed to work the exact same way every time (stupid human programmers!) If its very intermittant, and a reboot solves the issue, my thoughts would lean toward some sort of driver or protocol not initializing correctly.

Are all servers set to reboot at the same time? Are they staggered? If they all reboot at the same scheduled time, perhaps there's too much traffic trying to be put out from the farm at once?

Could i inquire though, why you are not running w2k SP4?

MIS Department
American Red Cross
Dallas Area Chapters

 

[rjs]

The reboots are staggered by 30 minutes. I found a bunch of messages in the Dell community forums about similar problems with the Netxtreme cards, so my inclination is to just buy a PCI NIC and disable the onboard BroadCom.

As for SP4, there were so many reported problems with it and Citrix/Terminal services that we initially held off. The system remains stable so we have never found a compelling reason to do SP4, though we continue to do all the security updates.

R.Sobelman

http://www.tracklogins.com

 

[oisnds]

Some NICs have a power saving feature that will turn them off after a period of inactivity. Strangely enough this feature is usually on by default. Check the properties of the NICs, should be on the Power Management tab.

 

[rjs]

Thanks, the power saving was the first thing I checked.

R.Sobelman

http://www.tracklogins.com