12-28-2014 05:59 PM
So I am in the process of building a new vmware cluster to move our production environment into and am deploying VCS for the first time, trying to decide whether it is worth using a vlag from the ESXi hosts or just to leave it as failover without the LAG.
It is important to note that all storage is dedicated FC with its own switching, so this is a purely ethernet setup question.
Quick synopsis of the setup.
4 switch full mesh 6740 cluster in logical chassis mode (2 top of rack 2 in chassis).
IBM Flex Chassis with 10 ESXi hosts, 2x10G Ethernet nics per host, 1 per in-chassis switch.
All vlans will be presented to all hosts and any VM can freely migrate to any host.
Our current setup is actually very similar but obviously with no VCS. We have the 2 nics in active/standby for data traffic, managed by the vswitch. DRS traffic is configued the same as regular data traffic but the active/standby is reversed to essentially segregate them except in times of a fault.
So a few referneces such as
Which is nearly identical to our setup (SRM included), recommend a vlag from the hosts to the VCS fabric.
The question I suppose is that if we use a vlag, will we really gain much over allowing vmware to perform a failover? It is pretty much seamless as is with maybe 1 packet drop, and since there is no storage traffic the risk of disk corruption is non existent.
I do see that with a vlag we would probably have to implement some QoS to manage DRS rather than letting it be relatively unfettered as it is now because it would be shared with data traffic.
Are there other advantages/benefits to using a lag in this situation?
Also, as our setup is completely flat currently, with all vlans living everywhere, we use vmware standard switches not distributed, still cant see any reason to use vmware distributed switches given our infrastructure setup, any gain to using them? Or just adding additional management for no gain?
Comments and suggestions appreciated.