I'm trying to get vNIC working.
I'm about to give up to all limitations and unknown configuration settings for my equipment.
What I have:
- BladeCenter H
- 3 x HS22V with Emulex II Adv
- 2 x Virtual Fabric 10Gb Switch Modules "VFSM" (Bay 7 & 9) - Firmware 126.96.36.199
- 2 x RackSwitch G8264 - Firmware 188.8.131.52
I want my HS22V to have all 8 vNICs available for VMware ESXi 5.x, connecting to iSCSI via uplink from VFSM to G8264 where my iSCSI SAN is connected:
- 2 x vNIC.1 for Host Management
- 2 x vNIC.2 for iSCSI
- 2 x vNIC.3 for vMotion & FT
- 2 x vNIC.4 for VMguest LAN (many vlans)
I've run into more or less all the limitations there are:
- G8264 doesn't support CEE in stacking (or IPv6 but that is not that important now)
- VFSM doesn't support vNIC uplink-share or lacp key when stacked
From having both G8264 and VFSM in stacked mode, i've now got G8264 in ISL/vLAG mode and VFSM in standalone mode.
What I have (done):
- VFSM EXT5+EXT6 are configured for LACP towards my ISL-G8264 with my vlans defined (vlag enabled since one 10GbE goes to each G8264)
- My Emulex II Adv set to Virtual Fabric Mode in UEFI
- All four vNICs defined with bandwidth and enabled
- vnic uplink-share enabled
4 vNIC groups defined
- one for each index (1.1, 1.2, 1.3, 1.4) (member)
- group vlan set to 3001-3004
- lacp key set equal to for LACP for EXT5+EXT6. This key has identical value as it's LACP partner on my G8264 switches btw (think I read somewhere it should be identical)
- global 'cee enable' and 'cee iscsi enable'
What I've understood is that the packets from my ESX server going out from vNIC.1 with vlan 100 will get an outer vlan 3001 and be sent out over my LACP uplink to my G8264, still packed with an outer VLAN of vlan 3001.
How do I receive this on the G8264, how do I strip this off so my packet can continue its journey to the right destination on it's intended vlan?
Do I configure vmnic groups on my G8264? add vlan 3001 to the lacp ports just on g8264?
One of my requirements are to stay with my two uplinks from each VFSM and not buy more SFP+ transceivers, which I would have to if I went with stack on VFSM and/or non-shared uplinks.