Hi folks,
Bit confused...
says: "VMware support only one EtherChannel per vSwitch or vNetwork Distributed Switch (vDS)."
This has confused me... I would have presumed I would need one LAG to each physical host? Do I only need one LAG to all hosts (for NICs on the VLAN in question) and ESXi does some extra magic?
----
Another point - I'm also trying to figure out, on a Dell Powerconnect 6224 switch, which is the correct LAG hash algorithm. VMWare states it wants "IP Hash" which I take to mean hash(srcIP,dstIP) but the only optoins I have all mash some other variables in such as UDP/TCP port, or for non IP hash, MAC, VLAN and Ethertype.
Anyone have the correct magic for Powerconnect switches? Google is not my friend today ![]()
As always, many thanks in advance ![]()
Cheers
Tim
Check out the following networking diagrams, they may help you visualize some solutions;
http://vrif.blogspot.com/2011/10/vmware-vsphere-5-host-network-designs.html
Regards,
Paul
Thanks Paul,
http://vrif.blogspot.com/2011/10/vmware-vsphere-5-host-network-design-6.html
Is probably the closest to my layout, except (initially) I will have the management net and vMotion net on different pairs of NICs with untagged VLAN uplinks for simplicity (though I like the approach for vSwitch0 in the diagram).
Unfortunately it does not really help with my question of whether I need 1 LAG group for 3 hosts, or 3 LAG groups (1 per host).
The latter is more usual for regular linux hosts, but that bit I quote din the VMWare docs about there being only 1 LAG per dvSwitch permitted confused me. It doesn't make sense to me to try to span a LAG across several hosts...
Cheers
Tim
I would caution you around using vMotion and Management on the same VLAN. VMware recommends against this approach. vMotion traffic is not encrypted and in a segmented network VLAN tagging is the most secure approach.
You can have more then one vDS per host, you can have several in fact. So if you need more then one LACP simply provision a different vDS for each one.
There are not actually that many instances where using LACP (what you call LAG or Etherchannel on Cisco switches) is worth while. In the 8 NIC diagram you can see Etherchannel provisioned for the FT switch, but that is the only one.
If you have vMotion and Management on the same switch you will get no throughput benifit from LACP.
If you have VM Networking hosted on a vDS then the best throughput will be gained via "Route based on physical NIC load" and will not benefit from LACP.
Does this help?
Regards,
Paul
Hi,
No, I was not going to put managemnet and vMotion on the same VLAN.
I *was* going to send them up seperate untagged uplinks to seperate VLANS on the switch. I only noted, with curiousity the tagged VLAN approach in that diagram. But I have plenty of switch capacity and plenty of NICs so I favour the KISS approach ![]()
Cheers
Tim
