How to setup vcenter 6 appliance high availability

BlizzardX

BlizzardX

BlizzardX

Soldato
Joined
19 Oct 2002
Posts
2,714
Location
Auckland, New Zealand
Hi,

I'm looking for some advice for a quick test. I've got the vcenter 6 appliance iso which I can install but I'd like to get a high availability setup done using my load balancers so that my single virtual IP can move between the appliances, depending on the hosts.

My hardware doesn't have the capacity to do the shared storage/10GB network for auto-restarting on another machine so I'd like to get two copies up and running at the same time and use a load balancer to get access to either of them.

I presume this can be done? If so, can I create two single instances that are joined together (i.e. create the first, then install the second and join in to the first), create the VIP on the load balancer and then add the real ips to that? Is there anything more to it than that?

Any pointers would be great. I am only testing this at the moment, and this is not for a production system at all.
 
Ok so it was a simple as installing two instances, joining them both together and then setting up the loadbalancers to a VIP connected to both servers... Easy.

VCSA seems to get a little confused if you're trying to connect while one has just shut down (or crashed); will play a little more with this.

Does VCSA 6 have a backend check script that my loadbalancers can check against to see if it is still up?
 
As far as I'm aware that's not a supported configuration. You don't need HA for vCenter, as the infrastructure doesn't rely on it to run.
 
Ok, I thought it was... How are you supposed to setup a failover to another host if the main host dies if you don't have the hardware to support fault tolerance that I believe is the recommended option?
 
You need to be careful with your terms. The VCSA is actually a couple of systems, new to v6. You've got the Platform Services Controller (PSC) which CAN be load balanced and the Virtual Center, which cannot.

You setup the PSCs as a LB pair, then point the VC install at the PSC VIP. If your VC fails for whatever reason, you restore from backup or create a new VC and reload the hosts into the new one.

Be careful with your load balancers if you're doing this in production, however - only certain ones are supported! http://kb.vmware.com/selfservice/se...ype=kc&docTypeID=DT_KB_1_1&externalId=2112736

And a nice howto on how to do it all here:
http://kb.vmware.com/selfservice/mi...nguage=en_US&cmd=displayKC&externalId=2113315

Having done this just recently, it all works nicely as long as you follow it carefully and have all your DNS entries setup ready :)
 
BlizzardX said:
Ok, I thought it was... How are you supposed to setup a failover to another host if the main host dies if you don't have the hardware to support fault tolerance that I believe is the recommended option?
Click to expand...

High Availability clusters do not need the virtual center to be up and running to allow VMs to be restarted on an alternative host in the event of a host failure; when you create a cluster and enable HA option FDM is installed onto each host. A master node is selected and the cluster is up and running via FDM agents on each node, the VC does not part take in the HA mechanism outside of administration.
 
ecksmen said:
High Availability clusters do not need the virtual center to be up and running to allow VMs to be restarted on an alternative host in the event of a host failure; when you create a cluster and enable HA option FDM is installed onto each host. A master node is selected and the cluster is up and running via FDM agents on each node, the VC does not part take in the HA mechanism outside of administration.
Click to expand...

Hi and thanks, yeah I tried that but my hardware doesn't have enough network ports or the 10gb requirement for updating the secondary held vcsa instance.. If it can work fine over 1gb then great but I wasn't sure, hence my thought around load balancers as that is how I run the test exchange 2016 servers.

Actually, I've just re-read this and I think I missed what you where saying in my response above. In this case, I'm only after high availability for the VCSA, not other VMs... I run a large amount of passthrough devices so each VM must remain on the host, even if it is down... My comments do still stand when talking about the appliance high availability, not just the hypervisor.
 
Last edited:
Dogers said:
You need to be careful with your terms. The VCSA is actually a couple of systems, new to v6. You've got the Platform Services Controller (PSC) which CAN be load balanced and the Virtual Center, which cannot.

You setup the PSCs as a LB pair, then point the VC install at the PSC VIP. If your VC fails for whatever reason, you restore from backup or create a new VC and reload the hosts into the new one.

Be careful with your load balancers if you're doing this in production, however - only certain ones are supported! http://kb.vmware.com/selfservice/se...ype=kc&docTypeID=DT_KB_1_1&externalId=2112736

And a nice howto on how to do it all here:
http://kb.vmware.com/selfservice/mi...nguage=en_US&cmd=displayKC&externalId=2113315

Having done this just recently, it all works nicely as long as you follow it carefully and have all your DNS entries setup ready :)
Click to expand...

Yes, sorry I understand that it is two components. So, I know that the PSCs automatically can load balance (well I think they can from all the diagrams I have seen) but you cannot load balance the actual VC server at all? I'd hoped it would have been like Exchange 2013 with the mailbox and CAS server roles where the mailboxes automatically talked to each other in the cluster and you just load balanced to the CAS servers...
 
BlizzardX said:
my hardware doesn't have enough network ports or the 10gb requirement
Click to expand...

Sounds like you're talking about FT there, rather than HA? HA requires no constant traffic beyond a heartbeat..

And yes, no VC load balancing - each VC has its own database so which doesn't replicate :(
 
Dogers said:
Sounds like you're talking about FT there, rather than HA? HA requires no constant traffic beyond a heartbeat..

And yes, no VC load balancing - each VC has its own database so which doesn't replicate :(
Click to expand...

Sorry yes I was, VCSA can run in FT mode where by it keeps a constant secondary copy of itself that is constantly replicated; this is the ideal one as it switches as soon as the master drops out. This is what I was trying to describe. Can this be done without a 10GB network connection?
 
The point is: vCenter doesn't need to be running for everything to still work fine. As long as you are running vCenter as a VM, you get hardware HA, and as long as you have a database backup, you could get a new vCenter running in a few hours. Or if worse came to worse, just install a clean vCenter and take over the existing hosts and reconfigure from scratch. There usually isn't that much to configure, so documenting your settings would go a long way to mitigating any issues. And again, a lot of these fine tune settings are more "nice to have" than essential.

Clustering and load balancing are incredibly hard to get right, and for once, VMware has made the right call by not putting together some janky solution that imposes a bunch of restrictions, introduces a load of complexity, and doesn't actually give you that much in return.

In IT, complexity is the enemy of stability.
 
I think it is a common misconception that people have, where they don't realize that vcenter can be rebooted during the day without any problems. It is only realy backup that ties in to vcenter that would be affected during a reboot. Everything else is either at the host or vm or os level.
 
BlizzardX said:
Sorry yes I was, VCSA can run in FT mode where by it keeps a constant secondary copy of itself that is constantly replicated; this is the ideal one as it switches as soon as the master drops out. This is what I was trying to describe. Can this be done without a 10GB network connection?
Click to expand...

It can - they strongly recommend you dedicate a physical gigabit or better connection for it though. How well it'll work over 1GBE I don't know though :)

Apparently there's a formula you can use:
VMware FT logging bandwidth = (Avg disk reads (MB/s) × 8 + Avg network input (Mbps)) × 1.2 [20% headroom]
 
Hi all,

Thanks for the replies, I appreciate that the host controls HA/FT, wasn't sure how vcenter integrated with it fully.

I can confirm that HA/FT do run on 1GBE network; seems to switch fine and is replicating at just over 100MB/s; so I suspect that I am the limit of the 1GBE network... although not far from the max rate of this test system as I've not set up RAID on any of these drives... Just testing as I said.

I presume if FT was NOT enabled, HA would just attempt to restart the VM on a host with enough resources, as long as the VM was on shared storage?
 
I can understand if you are doing this to play around with FT, but it is 100% completely and utterly pointless for you to do this in a real environment.
 
I am just playing around really, I was curious to know how firstly HA worked, then FT but also was curious whether you could load balance the VC itself.
 
You must log in or register to reply here.
Back
Top Bottom