BL460G6 network issue

[MKemp_1]

I have a c7000 chassis with 3120 switches set up as idependant switches for an A and B fabric. I have my BL460G6 servers teamed on the embedded NICs. The NICs are configured for 1GB/Full as are the ports on the switches. The ports on the switches regularly go offline and I have to disable them on the switch and re-enable them to get them back. Does anyone know what could be causing this? It is getting very critical that i get this resolved or replace the servers. There are 6 BL480G1s occupying other slots in this chassis and none of them are having this issue. The switches are running Cisco IOS 12.2(46)SE, all firmware and drivers on the servers are up to date as are the OA and 9124e switches.

[HEM_2 · Accepted solution]

Cisco IOS 12.2(50)SE3 for the 3120 should resolve the problem.

[MKemp_1]

Is this a known issue that is documented somewhere? The reason i ask is because I would have to upgrade my entire Blade environment to accomodate the new blades and I have 8 c7000 chassis, 6 with 3120x switches and 2 with 3020 switches.

[Pieter 't Hart]

>>>
independant switches for an A and B fabric. I have my BL460G6 servers teamed on the embedded NICs
<<<

maybe i read your post in another way you intend?
normally fabric is used in SAN environment, NIC is used with LAN, but I assume the term fabric denotes two isolated networks here?

If you have two switches not connected to the same LAN, then I think teaming the NICs on the server accross those switches puts you into trouble.
your network-team (same network) connects to two fabrics (separate networks).
when a host selects the team-member on fabric-B the nic on fabric-A get switched off.

[MKemp_1]

Yes the term fabric is being used to denote 2 separate LANs from the blade perspective. Of course they are connected at the core level and both LAN segments have all of the VLANs available to them. I have existing, non G6, servers on this configuration that are working just fine. It is only the BL460G6 servers, with the Flex-10 NICs, that are having this issue.

[Woosha]

This may apply to you as I had a similar problem with 10GBe Nics on BL460 G6 servers connected to 3020 switches.

This only happens with NC532i Dual Port Flex-10 and IOS version 12.2(46)or lower. When using a server with Multifunction Gigabit Server Adapters or a Switch with IOS version 12.2(50), the issue will not appear (There are no fixes related on the 12.2(50) IOS Release notes). As a workaround -

set the Link-Flap trigger values to a higher number using the following command:
- Default Settings:
Switch#sho errdisable flap-values

ErrDisable Reason Flaps Time(sec)
----------------- ------ ----------
pagp-flap 3 30
dtp-flap 3 30
link-flap 5 10
- Changing Settings:
Switch(config)#errdisable flap-setting cause link-flap max-flaps 10 time 20
- New Settings:
Switch#sho errdisable flap-values

ErrDisable Reason Flaps Time(sec)
----------------- ------ ----------
pagp-flap 3 30
dtp-flap 3 30
link-flap 10 20
* Note that this is a global setting on the switch itself, not on an individual interface