VLAN communication problem in ESX / FlexFabric combination

[chuckk281]

Johan had an ESX VMotion and Virtual Connect FlexFabric issue:

****************************

Hi,

I don´t know if this is a VMware or Virtual Connect issue so I send my question to both groups 

I have a problem with a Matrix Operating Environment customer where their Vmotion network  doesn´t communicate properly.

Servers in the same enclosure can´t communicate over this vNet, it is applied to FlexNIC 5 and 6 in the ESX-profiles in VC. They can´t ping or vmkping each other or their gateway with the VLAN id set to 617 in the portgroup in ESX.

If I check the “Native” checkbox below and remove the VLAN-tagging on the portgroup in VMware as shown below it works.

The VC fw is 3.17 and they run ESXi 4.1 U1 on BL465 and BL460 G7s.

**********************

Here is the conversation between Johan and Andrea:

*********************

1. <A>It seems that the packet coming from the SUS does not contain the TAG ID for the vmotion vNet/VLAN.

Putting it “native” simply eliminates the TAG and so the untagged portgroup can communicate with external network.

<JF> I know, that is why I removed the VLAN id from the portgroup, just to test if it worked. The problem is that even if we have two ESX-servers in THE SAME ENCLOSURE the error still occurs, so there is no external network involved.

<A> ok, and if you create an internal Ethernet Network called “vmotion2” those ESXi can vmkping each others?

2. <A>Is the VLAN tag the same for server and SUS?

<JF> Yes, it is set to 617. I have verified this myself and tested numerous times.

<A> ok

3. <A> ex. It has been used the “force same SUS vlan tags” option during VLAN creation in SUS?

<JF> No we didn´t use “Force same SUS VLAN tags” because the customer don´t want all of the VLANs in the SUS on their ESX-profiles.

<A> This is not a problem, you can select “Force same VLAN tags” and assign only a part of all SUS VLANs.

Are you assigning the vmotion network using a “Multiple Network” tag in the profile?
Even more: you can consider using fw 3.18, since I see many vlans in this configuration and if you remember latest advisories 3.17 does not support more than 70 vlans.

This normally happens when the customer changes the VLAN id after the VC configuration when VLANs were configured manually.

********************

Any other comments or questions??