HP Virtual Connect Stacking Failover (Multiple networks ethernet profile)
So I am currently on a project and one part of the hardware used is HP blades (2 chassis) with 2 virtual connect modules on each chassis, you can check here the Virtual Connect Multi-Enclosure Stacking Reference Guide.
In each blade profile we have three types of networks, the first two are single networks and the third is a multiple network (around 10+ VLANs).
We have two SUSs, one from chassis 0 and another one from chassis 1, so as per the above each blade server should have 6 interfaces.
When we came to the failover tests we faced an issue where when we powered off a VC module for example in chassis 0:
- All communication from/to the ethernet adapters belonging to the single network configuration were fine and 100% functional and the adapter belonging to the failing VC is no longer linked and down as well.
- All communication from/to the ethernet adapters belonging to the multiple network configuration was NOT fine and it was intermittent disconnects all the way, plus the adapters were still showing that they were linked.
On the single network configuration we had smart link checked, on the multiple network configuration we had smart link checked on a single VLAN.
Most of the HP documentation states that smart link should be enabled, the problem here was they it was enabled on a VLAN part of the multiple ethernet network and it was not working as it is supposed to.
After a lot of trial and error attempts we enabled smart link on all the VLANs part of the multiple ethernet network and only then it seems that smart link kicked in when doing our failover tests.
(Abdullah)^2
This could be due to the way SmartLink is designed. SmartLink will only kick in if ALL uplinks connected to a downstream switch port fail will the Nic be marked as down. So for example if you are using mapped VLANs and have connections up from either a single chassis or multiple chassis and have multiple uplinks, Say Port 1 going to SUS with VLAN 1,2 Port 2 going to SUS with VLAN 21,22. You have a server profile with VLANs 1,2,21,22 going to a blade. If the uplink from port 1 fails but the uplinks from port 2 are still working, VLANs 1,2 will stop working but not fail over as SmartLink won’t shut down the blade port as it still sees live connections from VLANs 21,22.
No way round it othger than to have only single uplinks going to a blade.
:-(
Indeed this is by design and I don’t have a document that says it is by design sadly :/. I figured it out when doing the high availability tests at the customer site, but now we know because I blogged about it :-P.
Thank you for sharing your thoughts and knowledge.