VCSA HA Stuff
There's already loads of great posts out there on the VCSA 6.5 HA option but I wanted to cover three things here:
- DNS issues with vCenter HA setup
- Failover Times
- vCenter HA Footprint
When I tried to deploy HA for my Lab VCSA it gave me an error message:
"Failed to get management network information. Verify if management interface (NIC0) is configured correctly and is reachable"
Check this Forum post for a fix:
https://communities.vmware.com/thread/547117?start=0&tstart=0
Basically you should ensure your vCenter name and DNS records are in the SAME CASE. I had my vCenter called "Labvc.lab.local" so I'd to edit /etc/vmware/systemname_info.json on the VCSA appliance to update the name to "labvc.lab.local" and then delete / recreate my DNS records to match the same case. After rebooting the VCSA I was able to deploy HA. Not critical but annoying!
Note: You will need three Hosts in your Lab (no less)!
From boot my Lab VCSA VM takes 5 minutes to boot until the old Web Client has initialized and is ready for logging on.
The vCenter HA Failover feature when initiated manually takes 6.5 minutes to perform a failover until the web interfaces are ready for logging on. (The newer UI is ready 20 seconds or so sooner than the older flash based legacy web interface).
So, would I use traditional VMware HA to recover vCenter or deploy this vCenter HA to perform a service failover instead? With an embedded PSC, vCenter HA has it's merits, it's only three VMs. If you are using it for VDI or other heavily vCenter dependant services it could be of use.
Once you get to an external PSC and load balancers, I think it's too complicated. Maybe with the next version of vCenter VCSA it might improve but 7 VMs for vCenter:
- 2 x PSCs
- 2 x Load Balancers
- 3 x vCenter VMs (A/P/W)
The old VMware HA to protect against host failures still delivers and can recover a VCSA in less time. Postgres corruption will still cause both VCSAs Active & Passive, to fail. A good VM backup strategy while using the inbuilt VCSA backup option should provide sufficient recovery options for all scenarios. If you need to go beyond this then vCenter HA is the next obvious choice but you had better have a load balancer or three handy!vCenter HA with two PSCs using manual repointing of vCenter still presents potential downtime between discover and remediation.
As for footprint, this is for a new build with no data or significant inventory:
So, some things to think about anyway. It's always good to have options and it's better / cheaper than the previous Heartbeat solution. If you're using Enhanced Linked Mode / have scaling requirements, then the number of VMs makes me think twice.....
These are just my thoughts, so evaluate for yourself and your environment before coming to any conclusions!
