Anyone used the Emulex OCE11102 card?

[JoJoGabor]

Interested in any experiences with reliability or anything else good or bad

Thanks in advance

[peetz]

We are using the HP branded version of this card (NC553i) with ProLiant BL620c G7 blades. In the beginning there have been firmware issues with both vSphere 4.1 and 5.0, but this is all fixed now. See http://v-front.blogspot.com/2011/10/update-esxi-50-on-hp-g7-blades-now-go.html.

- Andreas

[JoJoGabor]

Hi Peetz,

I am now testing this card and hae a few teething issues, wonder if you have the same:

I cant view NIC performance information within vcenter for the Network channels of UMC

With 2x UMC cards in a server and a vswitch using 2 UMC-NICs, failover works fine, but failback results in VMs being unavailable for 15 - 20 seconds, with inbound and outbound pings failing. Moving to 1Gbps copper NICs works fine.

Bandwidth allocations dont seem to be a maximum. If unused bandwidth exists, it will be consumed regardless of allocation.

It could be something in my setup, I am using ESXi5 Vcenter 5, latest Emulex CIM provider, but interested if you've seen any of these issues in your environment?

[peetz]

I'm using the HP branded Emulex OneConnect adapters in HP c7000 blade enclosures and FlexFabric Virtual Connect modules. NIC partitioning is done through VirtualConnect/FlexNICs. I don't know if this is similar to UMC.

I do not experience the missing NIC performance counters with this configuration, but the failback issue is the same for us.

The failback issue can be solved by configuring the vSwitch teaming policy with "Failback = No". Probably failback occurs too fast in your environment when a NIC recovers, but is not yet ready to forward traffic. With Virtual Connect configurations there is a similar issue, so we also have configured our vSwitches with Failback=No.

- Andreas

[JoJoGabor]

Interesting, sounds like your cards are different, but same symptoms. Have you tested using dvswitches? Youre right in setting failbak to no will work round the problem but will result in unbalanced NIC utilisation. I am thinking if I use dvswitches, set the equivlanet of failback=no then let the dvswitch in ESXi5 take care of load balancing.

[peetz]

Yes, this is exactly what we are doing: Using dvSwitches with the Load Balancing policy "Route based on physical NIC load". This will spread VMs over all physical uplinks again after a while.

[elxaamaya]

JoJoGabor,

The OCe11102 adapter is a card for rack mount servers that may have similar characteristics as the 553i but not 100% exactly. There have been issues reported with failover and failback on the OCe11102 enable with UMC. I suggest before contacting Emulex support check to make sure you have the latest FW (4.1.402.18) and corresponding NIC/iSCSI driver (4.1.334.0) installed for ESXi 5. if the problems still persist contact Emulex support and have them verify the issues you are encountering and possibly add you to an incident report or it might be already in the works to be addressed with a new FW and driver.

--Alex

http://www.implementerslab.com