VMWare vSphere Cloning/Management Network

blueboy2001

blueboy2001

blueboy2001

Soldato
Joined
18 Oct 2002
Posts
4,898
I'm trying to troubleshoot an issue that when we clone a VM in VMWare vCenter 5.1, it causes our LAN to grind to a halt - despite all the storage being on a separate 10GbE network.

I don't know much about the mechanics of vSphere, but can anyone confirm something I've read tonight that when you clone a VM all the data runs over the management network for the cluster?

If that's right, it explains the problem. I think the management network runs over the LAN switches. Does the management network need access to anything other than the hosts?
 
When you migrate or clone a VM the data moves across whichever management network you have configured for vMotion.

What you might want to consider is leaving your existing management network on your LAN but adding in some additional networking specifically for vMotion, probably 10GbE or better. Your current management network needs to be accessible from all of your hosts and the gateway on that particular subnet, but not necessarily from anything else other than your vCenter server (provided you have a means of reaching vCenter externally from the management network).

I think that's right anyway. Chip in if I'm wrong please :)
 
Sounds like you have Vmotion on the same network as your management network.
Cloning a VM is just not very network intensive, Vmotion is.
We run our Vmotion on a separate VLAN on the Gbe switch.
Someone may have the correct answer.
 
Cloning runs over your Vmotion network, as already stated.

Really its still pulling up from your storage network, and writing back to it. If you clone straight away between host-to-host then yes that Vmotion/management network gets used.

That said, both our 'front-end' productions networks and our SAN Storage platform run on seperate 10Gb switching, with the correct jumbo configuration throughout, so cloning takes minutes for us.

We use a Dell 8164 PowerConnect on the front-end, and a Dell 8024 PowerConnect switch on the back-end.
 
Or you can use VAAI enabled storage, which will perform the clone for you in the back-end (without having to stream the data up and back down).
 
Thanks guys. We've had further problems this week with everything grinding to a halt during normal day to day operations.

We might have been barking up the wrong tree with the cloning issue being related to the network. vMotion is checked on the 2x 10GbE NIC team used for the storage connection - this is on a separate set of switches to the LAN. Whilst I'm not convinced this is a good configuration as it could cause contention between storage and vmotion, it shouldn't be hammering the LAN. I've ordered 2 new ESX hosts with 4 10GbE ports, and an extra 10GbE card for each of our existing boxes and we'll have vmotion on a separate vlan with separate interfaces.

The problem appears to stem from our storage. We had 2 NFS datastores (named capacity and performance), one running on a big aggregate using 48 15k SAS disks and another running on another aggregate with 12x 2TB SATA disks. The SATA disks also serve CIFS for 600 users home drives, profiles, department shares etc. We've ended up with 30 VM's running on capacity, plus its used for our VM templates and ISO store. The disks were maxing out and causing huge latency, which was causing timeouts all over the place. I suspect the same issue has been occuring during cloning.

We're moving all the VM's off SATA for now.
 
By the sounds of things your NFS datastores were whitebox roll-your-own types since you didn't say NetApp, so you're missing out on all the VAAI goodness.
 
You must log in or register to reply here.
Back
Top Bottom