Making vcenter high availability

[legopc]

I am currently looking for a way to make vcenter highly available. Every single time vcenter has fails it has been an application error or windows shitting itself and making the VM unusable, so every time I have had to reinstall and begin from scratch. I am looking to make it HA as in one VM fails and the other one takes over, not as in 2 hosts and then have the VM switch host. I am not at all worried about hardware failure.
I have been looking into ways to do this but I have not found a lot of information about it. Unfortunately 1 ESXI host can not be managed by 2 vcenter servers, otherwise I would have been done already. The one thing that I have found that I think I can figure out is using Microsoft clustering (MSCS), but there is barely any documentation or proper guides on how to make it work. So I was wondering if anyone knows any alternative ways to achieve what I want to do or if anyone has any proper start to finish guides.

 

[markpower28]

sounds like vCenter appliance is your best option.

 

[legopc]

sounds like vCenter appliance is your best option.

What do you mean?

 

[markpower28]

vCSA (vCenter Server Appliance) is a virtual appliance for vCenter. It not windows based and MS SQL is not required.

 

[whitey]

Maybe a case for FT (fault tolerance). With vSphere 6 you can go multi vCPU/SMP FT, up to 4 vCPU's. I know VMware retired vCenter heartbeat so no legs there.

 

[legopc]

vCSA (vCenter Server Appliance) is a virtual appliance for vCenter. It not windows based and MS SQL is not required.

Hmm yes, I do prefer windows (linux hates me) but it could be something to look into. Thanks

 

[markpower28]

or another ways is create another management cluster and put vCenter VM in there. Make sure have good backup and snapshot.

 

[CreoleLakerFan]

Hmm yes, I do prefer windows (linux hates me) but it could be something to look into. Thanks

It's a virtual appliance, so there is little to no Linux admin involved. Everything is configured via a web interface.

 

[modder man]

What creole said the appliance just works, there really is no excuse to not use it.

 

[whitey]

Thirded, VCSA FTW!

 

[legopc]

I did manage to get the windows cluster working after many hours of work . But one thing I did find is that proper documentation or guides seem to be basically non existent, so I might write a guide and put it under the resources for other people. Next time vcenter fails one me again I will just switch the VCSA.

 

[EffrafaxOfWug]

Yup, steer clear of trying to do any funky clustering or HA functions for virtual centre; they just don't work, at least not in a "set it and forget it" sense. They're certainly more trouble than they're worth and unless you want to spend a significant amount of time on admin, the appliance is probably the best bet. It's based on linux and postgres and is reasonably bulletproof. We basically rely on a VC in each datacentre in linked mode and the ability to quickly restore an image-based backup in the event of an OS or application clusterfuck; it works well and means you'll be frequently checking the integrity of your backups which is a good habit to get into

Maybe a case for FT (fault tolerance). With vSphere 6 you can go multi vCPU/SMP FT, up to 4 vCPU's.

FT for multi cores has pretty serious bandwidth requirements though, I think 10GbE for the heartbeat is a minimum. It only really protects against host failure anyway - if it's an application or OS error that blows up your VM, the other VM in lockstep will almost certainly have the exact same issue at the exact same time. We evaluated it and found it pretty pointless for our workloads.

 

[legopc]

Yup, steer clear of trying to do any funky clustering or HA functions for virtual centre; they just don't work, at least not in a "set it and forget it" sense. They're certainly more trouble than they're worth and unless you want to spend a significant amount of time on admin, the appliance is probably the best bet. It's based on linux and postgres and is reasonably bulletproof. We basically rely on a VC in each datacentre in linked mode and the ability to quickly restore an image-based backup in the event of an OS or application clusterfuck; it works well and means you'll be frequently checking the integrity of your backups which is a good habit to get into



FT for multi cores has pretty serious bandwidth requirements though, I think 10GbE for the heartbeat is a minimum. It only really protects against host failure anyway - if it's an application or OS error that blows up your VM, the other VM in lockstep will almost certainly have the exact same issue at the exact same time. We evaluated it and found it pretty pointless for our workloads.

Yeah, your right. Clusters dont really work that well with vcenter I found out... All went fine for a little while, I tried shutting down one node and seeing how it would do and it went alright but when I tried a couple more times today it basically yet again shit itself... I will continue to attempt and cluster vcenter using MSCS for shits and giggles and for learning experience but I wont use it to actually manage the servers. I think I will be installing the VCSA in linked mode and figuring out how that works.