network problems on virtual servers

[zvitins]

Hi all,

I have vCenter 4.1 and 4 hosts (ESX and ESXi 4.1) with several virtual servers on each host + FC as data-store

On each host are standard switch with several vm networks - all hosts are in HA cluster.

On each switch are two nic in teaming mode.

When I tried migrate (using vMotion) virtual servers form one host to another I lost network connectivity for virtual server - I disconnect network adapter in vm settings and connect it back - then all is fine again.

I tried change vm network to different network and then back - also helps.

I tried this migration lot of times and this problem are in most of cases.

In vCenter Evens are no event about this problem.

I can not understand where to look for problem - any suggestions?

Thanks

Zigmunds

[a_p_]

When you say "teaming mode", I assume you connected two uplinks and run the default policy "Route based on originating port id".

The issue you see is most likely caused by the physical switch port configuration. Either you did not configure "spanning-tree portfast" or the port is not set to "mode access" and therefore limited to allow only two MAC addresses (Port security enabled/Macro "Desktop").

André

[cmacmillan]

Zigmunds,

Check your vSwitch teaming type. The default is based on "port ID" and requires ONLY that the VLAN configuration on the physical switch matches the VLAN configurations in your port group. While it is a good practice to enable "portfast" on Cisco switch ports (and similar options on competing switches) this has no effect unless you encounter physical port issues. Likewise, spanning tree should be disabled on all physical switch ports connected to a ESX host.

If you are  trying to use "IP hash" teaming on the ESX vSwitch, you need to enable "static" port aggregation groups on the physical switch. Please note that not all physical switches can span aggregated ports across phyisical switches so using the default teaming will be necessary in those circumstances.

Some switches will see MAC address movement as flap. Check you switch logs and take the vendor appropriate action to remedy those issues.

[zvitins]

Hello all,

yes both adapters are in active state with Load balancing "Route based on the originating virtual port id"

My hosts are on two blade system and all ports on the blade switch are configured as trunk with several vlan id (same vlan id are on vSwitch's)

Should I configure any spanning-tree options on the blade switch ports?

I tried another migration process and on the switch logs are no entry about any problem in this time.

how can I see logs from vSwitch?

Thanks.

Zigmunds

[kjb007 · Accepted solution]

There are no vSwitch logs, per se.  As the vSwitch does not participate in spanning tree, spanning tree options typically do not matter with ESX hosts.  Getting your network team to diable it all together is typically not well received.

Portfast, however, should be enabled as posted above.  You'll need to make sure portfast is enabled on your access layer switch.  If they are in the blade system, then you need to make sure portfast is turned on at that level.

After a migration, without portfast, the migrated nic should eventually come up, and you should be able to see the mac address in the appropriate switch.  Are you able to see that?

-KjB

[zvitins]

Hi,

thanks, I will try your suggestions.

Zigmunds

[Josh26]

Kanuj Behl wrote:

Getting your network team to diable it all together is typically not well received.

For good reason.

Disabling spanning tree then letting a loop occur is a recipe for disaster.