c7000 BladeSystem interconnect problems

[mosc9562]

Hello,
We have a c7000 bladesystem, we originally purchased 3 half height BL460c G1 servers, which are working fine. We have now purchased several other half height servers, BL460c G1's, and we are running into network issues. Seems that when I look at Port Mapping in the HP Onboard Administrator for the servers or the switch (we are using two GbE2c switches) the new servers Embedded Ethernet Device ID's are all 0s. The original 3 servers have the mac address but for some reason the new servers device ID's are not showing up. Anyone help me with this?? As a result their network connect is not stable, it will be working fine for a little while and just stop. Thanks.

[Change_happens]

See here port mapping:
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1222578

[HEM_2]

the display of MAC Addresses in the OA shouldn't be too relevant to the actual connectivity of the servers.

I would ensure that you have latest system ROMs, ILO firmware and OA firmware to potentially clear up the display issue.

Connectivity issues can be from many different causes.

One of the most common currently, that I would recommend looking in to is if you are running Windows 2003 SP2:

http://support.microsoft.com/kb/948496

[mosc9562]

Thanks for the info. Last night I went ahead and updated the ILO and onboard administrator firmware to the latest version, and that did clear up the device ID problem. But my overall problem is still there, that the new servers loose network connectivity. Example, if I start a ping and let it run to any of the new servers it will eventually time out for a bit and then start back up. This does NOT happen on my 3 original servers, only the new ones. Also, all servers have the Proliant Support Pacs 7.90 installed and have Windows SP2 installed as well.

Any other ideas?? Thanks.

[Igor Karasik]

>> Example, if I start a ping and let it run to any of the new servers it will eventually time out for a bit and then start back up.

I saw same behaviour on some our ML servers (not blades) with windows 2003 sp2 and only 1 Gb of RAM and disabling RSS/TCP Offload help us. This not happens on ML with 4 Gb RAM, so I think it related to TCP offload memory issues.