DRS and HA Config Suggestions Please

[chris_delaney]

Dear All,

I'm planning to implement DRS and HA in my VI3. I have 3 ESX hosts in a 'production' site and a single ESX host in a 'DR' site. Ideally I'd like to have a DRS cluster that moves VMs amongst the three production ESX hosts. Along with this I'd like an HA cluster that can move VMs to the DR host should anything happen to one of the production ones.

I assume that I'll need to add all four hosts to an HA cluster but is it possible to have just the three production hosts in the DRS cluster?

Many thanks.

Chris

[ThomasNederman]

It is not as easy as that. You need all hosts in the same cluster, you can not select a few hosts to be in a HA cluster and other hosts to be in a DRS cluster. The way to implement it is to use mirrorview or another replication tool to replicate the storage to your DR site. Then you probalby have to write some scripts that see if the origianl host goes down and turn on the DR host.. However, this requires changes of DNS probably, if the DR and production is not on the same subnet.

[chris_delaney]

I had a feeling that what you describe was probably the case. I am currently replicating the VM volumes between our production and DR SAN appliances (EqualLogic) so I could probably fiddle some Perl scripts to do a failover. Probably the easiest option at the moment would be for me to have all four hosts running in the DR and HA clusters and then having a separate procedure in case I lose the production site completely. It'll just be a case of pointing the DR host at the DR SAN and using the replicas I suppose.

Thanks for your input.

Chris

[ThomasNederman]

One way of doing this would be to make the VM take a snapshot, then copy the VMDK file to your DR, apply the snappshot to the VMDK file. Put this as a scheduled task.

The only problem with this is what happens if the crash happens while copying the file between prod and DR ? You need to temporary store the file under a different name during copy and then change name when you are sure the copy is ok.

have never done this myself. My instructor on my VCP course told me he done this at one of his clients

Thomas Nederman, VCP

http://www.thomasnederman.com

[chris_delaney]

I have written some Perl scripts that suspend, snapshot (EqualLogic volume snapshot) and replicate (EqualLogic volume replicate) certain VMs over to our DR site every evening. This seems to be working really well as the EqualLogic boxes do most of the work. Plus, of course, the EqualLogic snapshots only take a second to do and the replications are just a snapshot that gets moved over to the DR site. This being the case they are independant of VMWare and should be resistant against any problems therein.

The SAN and VMotion networks are connected to the DR site using MSTP trunks that should cope with pretty much anything bar total destruction of the production site. Perhaps site production manager will offer a cleaner solution - but at a price!

Cheers.

Chris