Portgroup/VLAN intermittent loss

[ronald_higgins]

Heh fellow VMware Admins,

Thanks for reading this so far :smileygrin:

I've got this really odd issue with ESX3.5 and Port Groups.

I'll create a VM and slot it into PortGroup named 'mail', mail's vlan id is 20.

I'll start up the VM and i'll have no network connectivity.

Shutdown the VM, edit Network Settings, change the portgroup to any other portgroup, SAVE. Edit again, change back to the mail Portgroup, Save and then restart, if i'm lucky, the VM will start in the right Portgroup and i'll have connectivity. Somtimes I have to

repeat this process a few times before it seems the VM get's stuck into the correct VLAN(PortGroup).

It's been a bit of an annoyance up until now, and i've just dealt with it for the last month. But we're now looking at using VMotion.

What happens now is, the VM runs fine, VMotion it too another host, and the connectivity is lost, as if it's lost the VLAN along the way.

Anyone seen this behaviour before?

Regards

rH

[depping]

Never seen this before, are you 100% sure all your portgroups have exactly the same name?

Duncan

Blogging: http://www.yellow-bricks.com

If you find this information useful, please award points for "correct" or "helpful".

[ronald_higgins]

Yep ...

100's sure.

The VMotion issue is over and above the first issue which I think really is related.

[depping]

It's definitely related. Can you find any messages in the log files related to this error?

Duncan

Blogging: http://www.yellow-bricks.com

If you find this information useful, please award points for "correct" or "helpful".

[Rubeck · Accepted solution]

I would verify that the VLAN is correctly trunked to all pNICs.... If one or more pNICs is missing the VLAN 20, you could experince this as you have no control on what adpater is selected for the VM at power-on.

I would check "oberserved IP ranges" for all pNICs to verify that VLAN 20 is seen by all of them...

/Rubeck

[ronald_higgins]

Thanks for all the feedback to date.

I've taken the VM out of service, as soon as the connections to it drop i'll VMotion it and tail the logs.

Rubeck, you may be onto something regarding the pNic's.And all of the sudden I get the feeling i've done something very silly 😛

Our ESX servers each have 2 pNICs. Each plugged into a different switch.

There is 1 Vswitch with both adapters. Both adapters are "Active" and only 1 of the adaptors is being advertised the required VLANs.But when failed over the required VLANs are advertised to the secondary NIC.

I'm guessing that as you suggested, when the VM is powered on the networking is binding to the 2nd adaptor too which the vlans are not being advertised.

Lemme test.

rH

[depping]

that's definitely causing this problem. you should advertise all vlan's to all nics in the same vswitch/portgroup.

Duncan

Blogging: http://www.yellow-bricks.com

If you find this information useful, please award points for "correct" or "helpful".

[ronald_higgins]

Thanks for the feedback,

Yep, I get that feeling as well.

I cant argue from the networking trunk point of view as that is not my area of expertise but I will take that up with the Networks Team.

I'll move the second adapter into the Standby field after Change Management procs and will definately provide feedback. That will only leave 1 Active adapter for the VM's network to bind on.

Will provide feedback soonest :smileygrin:

rH

[ronald_higgins]

Awesome stuff Gents,

U rock!

Moved the 2nd adaptor into standby, and it all works a treat now :smileygrin:

Happy Friday!