OA Active/Standby Transition

[René Falkenberg]

Hi

We have problem where we cannot Transition from the Active OA to the Standby OA without some of the servers in the Enclosure looses they network link.

Anyone seen similar beheavior before ?


HW Info:
685c G6/G7 blades.
OA Firmware 3.11
G7 - iLO3 Firmware 1.15
G6 - iLO2 Firmware 1.82
Cisco 3020 Switchs

[Ali]

Hi Rene,

what is the firmware of the standby OA?

they should be on the same level.
check the OA sys log to any error messages

thanks,
Aftab

[René Falkenberg]

Well both OA's are running 3.11

What we see is that everytime we switch the OA, the servers in bay 1,9 2,10, 3,11 (Fullsize) loose they network connectivity on gi0/1, gi0/2, gi0/9 for around 16-20 secs.

This then course our VMWare HA to see the system as isolated and begin to move it's guests.

It's only on the mentioned bays. The bays on the right side of those three are not having the same issue.

[The Brit]

Rene,
are you saying that your server blades iLO/Consoles lose connectivity to the outside network, or that they lose connectivity to the OA. As far as I am aware, the only communication that passes through the OA to the blades, is to the iLO's and Consoles.

The applications which run on your server blades communicate with the External network through whatever InterConnect devices you have installed.

A transition between the Active and Standby OA's, successful or otherwise, should not (in principle) affect the communication of your apps with the outside world.

(of course, many weird and wonderful things seem to occur with these chassis.)

Please confirm that you are not using the iLO IP addresses at the system NIC level.

I suspect that this could cause similar behaviour to what you describe.

Dave.

[pakm11]

I had similar issues with
OA @ 2.60
VC @ 1.40

Only full height servers (BL 680c G5) were affected.
Note there is no Cisco switch in the enclosure.
All Interconnects were VC (1-2 & 5-6 Ethernet).
With the active-->stdby transition, I'd loose both members of teamed NIC for full height 680c servers in bays 1 and 2.

I have upgraded the f/w to 9.20B and yet to retry OA failover again.
And yes, NICs in the team are mapped to two different VC modules: bay 1 and bay2.

[René Falkenberg]

Yes the blades loose connectivity just after the OA has finished it's transition. When i login into the enclosure just after the restart, i can see many of the blades is missing.

In the same second the blade show up in the enclosure, i see the "Network connectivity lost" in my VMWare log and if i have a ping runninng it looses the 3 echo's before resuming.

I have tried to update iLO to 2.05 on the G6 and 1.16 on the G7, and the failure is still present.

[SMR]

René,

It would be interesting to know if the NICs actually lose link, check the IML on the affected blades for messages related to Link loss, it would not hurt to physically reseat the OA tray.

[René Falkenberg]

None of the IML logs show any Network link failures.

[René Falkenberg]

Reseating the OA's did not help either.

But it did not cause network link loss either. So it's only the transition that is causing the network link loss.

[SMR]

You could (if possible) swap one OA module with one from another enclosure and test the failover again. That would help determine if one of the modules is the culprit of the behavior.


When I said reseat the OA tray I meant the tray that holds both modules together, not the OA modules themselves :)