ESXi 5.5 Dell M620 M8428-K

[Geoff_Rose]

Hi All,

I have an issue with ESXi 5.5 - when I join all my NICs to the one switch I lose connection.

Here is my setup:

Dell M620 blades in a M1000e chassis. We have 10GB Qlogic CNA adapters - dual channel QMD8262-K that are partitioned into 4 adapters on each channel, connecting to 2 x Dell M8428-K (rebadged brocade 8000) chassis switches.

When using ESXi 5.1 everything works perfectly.

I have had a case open with VMware since November and so far they have been able to identify that the ARP requests are going out vmnic0 but the ARP reply packets are coming back through vmnic2 and are dropped by Team_ReverseFilterPerList(). This is why vmk0 stops sending traffic.

Just wondering if anyone has a similar setup and has it working with ESXi 5.5? we have 7 of these enclosures and none work with ESXi 5.5.

Thanks,

Geoff.

[DavoudTeimouri]

Hi friend,

Let us know, what was your installation image? do you using customized ESXi?

Also please share your vSwitch configuration, load-balancing policy and etc.

[Geoff_Rose]

Hi Davoud,

I've tried both and updated NIC drivers too.

I've tried all but mostly I'm using route based on originating virtual port ID - which is working fine on 5.1

I've tried using standard and distributed switches and both have the same issue.

NOIC doesnt have any effect either.

If I reboot the Dell chassis switches it generally works again until a reboot or host networking change.

Quite a strange bug between 5.5 and the chassis switches I think.

Geoff.

[rambosuri]

Do you still have the issue or is it fixed? are you running your upstreams to 2 x cisco switches ? I am guessing you are using Qlogic FCOE cards in M620's ?

[Geoff_Rose]

I still have it - VMware have requested assistance from QLogic.

Do you mean from the Chassis to Cisco? Then yes we are.

Yes thats right - the QMD8262-K is a CNA

[rambosuri]

Are you using ( A1,B1 or C1 ) bays on the chassis ? What versions of Qlogic firmware and drivers you have in 8262 ? , Are you using any trunks to upstream switches ? Are you doing Active/Active LBT ( And are you using IP Hash or route based on originating virtual port)?

How are you mapping the actual NIC's ? Inside your teaming ? Like M620 ( dual port) assuming if its connected to B1 and B2 and B1 uptreams to Cisco1 and B2 upstreams to Cisco2 , are you getting NIC mappings from both of these and pairing them together ?

[kristenthorsen]

I'm also experiencing exactlu similar symptoms with the same hw setup and sw version. I'v filed support tickets to both dell and vmware. Still no game after a couple of months troubleshooting.

<!-hepp

[Geoff_Rose]

It took vmware 9 months to come back with a solution. There is a simple command that can be run to fix it. I'll post it here tomorrow.

[rambosuri]

What version of Distributed switch  are you guys running ?

[Geoff_Rose]

This is a 5.5 issue so with that version. Also the nexus 1000v had the same issue

[kristenthorsen]

The issue is oberved on both 5.1.0.000 and 5.5 versions of dvswitches on esxi5.5++

brgd

kristen

[Geoff_Rose · Accepted solution]

Here's the command that worked for me. You need to reboot after its entered for it to take effect.

esxcfg-module -s defq_filters_param=0 qlcnic

[rambosuri]

what does this parameter change exactly do?

[Geoff_Rose]

It changes how the Mac address are handled. Can't remember exactly what though.

[rambosuri]

Does this also fix the issue of Vmotion of VM across hosts , where some time a random VM looses network conectivity ? Or if you reboot VM some times it looses network connection , the only way to get them back up is to disconnect and re connect the adapter from VM edit - network ?

[Geoff_Rose]

I didnt have that issue - it sounds different to me