radditsu
Silver Knight of the Realm
- 4,676
- 826
You're vmotioning VMs between physical datacenters? Madness I say!
Anyway, you positive all the nodes you are using have 10g links on them, and the 10g links are the nics assigned for vmotion and management traffic? Our nodes for example have 10g pci nics in them, but that connects to storage only, they run over the onboard 1g for vmotion and management traffic etc, but none of that is leaving the confines of their DC so it never really matters. If you have a 10g link between these two DCs and it tests at 10g speeds outside of vmotion traffic on these nodes, occams razor says its the nodes obviously. You on the network team it sounds like?
Well we aren't exactly a huge operation, we have a failover rack in another location ~ a mile away. We have got the VM support involved. They can never figure out anything.
It is on a isolated management VLAN so ONLY layer 2, I got into the cli of the VM's yesterday and poked around and everything looked fine. Going into deep dives could it be some cache setting maybe? I was going to mirror the port today and see if there was some sort of MTU issue, but I find it hard to imagine a MTU causing 90% degradation in speed. Could it be a HDD write speed? IS there a way i could do an IPerf like test between ESXI management hosts in the CLI? So many questions. The more I dig into it the more I don't THINK its a network thing. But I don't want to be the asshole.
- 1