02-22-2012 12:09 PM - edited 02-22-2012 12:19 PM
Can someone explain to me the appropriate way to configure a fully redundant Active/Active SUS in VC when one than one enclosure is involved?
Am I missing something?
Per the latest FlexFabric Cookbook, I'm able to setup redundant Active/Active SUS with multiple VLANs.
However, when I've tried to stretch this configuration to multiple enclosures, I end up with half of my uplinks not showing correct connections.
2 x C7000 chassis, each one with 2 HP VC FlexFabric 10Gb/24-Port Modules (Bays #1 and #2.)
2 x HP 5406zl switches, each with 2 10GBE x8 SFP+ ports.
16 BL620s G7 and BL460G7 Blades.
Various others switches, including 2910AL with 10GBE modules, etc.
FCOE is configured, and seems to be working properly as configured.
For ethernet, I've setup two SUS (VLAN-Trunk1, VLAN-Trunk2), each one with similar networks associated with them.
VLAN-Trunk1: ClusterHeartBeat (VLAN 112)
VLAN-Trunk2: ClusterHeartBeat2 (VLAN 112)
Everything works fine with a single Enclosure:
The physical connections, ports X5 and X6 on each FlexFabric 10Gb/24 module are connected to two LACP trunked ports on each HP 5406zl switch.
encls_0:Bay1:X5 <-> 5406zl:A1
encls_0:Bay1:X6 <-> 5406zl:A2
encls_1:Bay1:X5 <-> 5406zl:B1
encls_1:Bay1:X6 <-> 5406zl:B2
STP, IRF is setup on the ZLs and they're linked via ISC.
I can get this same configuration to work on the second chassis, as long as they're in different VC domains and two chassis are not stacked.
However, since my use case is to create a stretch cluster (MSCS w/HyperV), I added both chassis to the same VC domain, and created a stacking link between the two Enclosures.
I followed the VC Multi Enclosure Stacking Guide, and connected port X1 on each FlexFabric 10Gb/24 module in the first enclosure to the same ports in the second enclosure.
encls_0:Bay1:X1 <-> encls_1:Bay1:X1
encls_0:Bay2:X1 <-> encls_1:Bay2:X1
Doing this causes the links (X5/X6) to go from Linked/Active & Linked/Active to Linked/Standby & Linked/Standby on one of the chassis. The stacking links are working fine.
But looking at the configuration, I no longer have true redundancy to each blade. Is it not possible to have this configuration?
BTW, I'm runnning VC firmware v3.51 on these modules.
What is the correct way to make sure that all blades have truly redundant Active/Active links to the network?
If you need me to diagram this setup for better clarity, please let me know.
02-22-2012 03:08 PM
it will be better if you can provide a screen capture for your SUS summary view or you can do "show uplinkset" and "show uplinkset <SUS name>" in CLI.
the key thing is that if you want all uplinks active, assuming you have 4 VC modules and you have LACP uplink bundle from each module, then you need to define 4 SUS cover these 4 bundles, one SUS per bundle. If anything you have a SUS covering uplinks from two different VC modules, some links will be put in standby by VC due to its loop control mechanism.
For some internal cluster network like heartbeat, if they don't need to go out, you can define a network without any uplink, every blade can use this internal VC network and it can communicate across enlosures without leaving VC uplinks.