LACP, LAG, iSCSI

[Ray786]

We have some hosts with two physical NICs, some as iSCSI offload capable i.e. they appear as storage adapters as well.

We'd like to use LACP/LAG to link the NICs together for normal VM traffic, port security is on so just binding the NICs will end up with the NIC down so LACP is required.

Is there a way to then split the iSCSI traffic back off the LAG and down individual NICs as obviously iSCSI over LACP is bad.

The method of overriding the redundancy for the iSCSI vmnic doesn't seem achievable as you can't separate a LAG like you can with dvuplinks.

Is this possible is there a walkthrough?

[vfk]

Hi,

You should approach differently; if you have enterprise+ license then you should not really bother with LACP that much, just use LBT and NOIC.  Keep your environment simple.

Secondly, what benefit will you get from implementing LACP/LAG?  What is the traffic flow between your VMs?  Most virtual environments don't really get much benefit from LACP.  And if you have 10Gbe NICs, you should not really be worrying about LACP in a virtual environment unless you have a valid reason and this should have been identified earlier on.

[Ray786]

It sort of stemmed around tidying up physical network ports s they currently have them as active/orphan, port security etc. but...

Like you say it's better to have ESX handle it, so I'd guess using LBT and NIOC you would still segregate your iSCSI traffic down the HBA enabled cards and LBT across the other two uplinks or just do it across the 4 uplinks and then use port binding for iSCSI down the dedicated NICs?

Also on the physical switch side the uplinks would all be independent active links no special config?