Hyper-V split brain. Possible to have second server to only start manualy?

[DCUK6]

Interested in Hyper-V and things it can do.

Havnt got a second server to play with it so thought I would ask.

 

 

If you had two server 2012 R2 Hyper-V servers and they were replicating for failover and the backup server looses connection to the main server, could you stop it auto firing up the second server.

 

Thinking, in a school having the network go down for 10mins whilst you log in and start up the second servers is better than having both on and then having to spend hours/days recovering from data being written to both.

 

 

Is there a way to stop the second server automatically starting if it looses connection to the first server.

 

Another thing ive just thought of is what do you then do if the second server has started and taken over running things and you then turn the first back on. Is the second server now the main server that you carry on using for everything until it fails and the original main server takes back over or will the original main server take back over at some point?

 

Just interested but don't have the hardware to test it.

 

Cheers.

[Steve21]

You can set the "starting" properties of a machine on failover. e.g. Don't start, start low, medium, high etc

 

But if you're doing just replication it shouldn't start anyway automatically.

 

Depends if you mean like cluster wise, or pure rep?

 

Generally we have the servers balanced between the two, and then "if" you want you can set them to auto shuffle between the two. Not using one as main and one as backup as you mentioned.

 

Steve

[Steve21]

To clarify my waffle:

 

(as far as I know)

Basic HyperV replication doesn't support automated failover.

You "can" script this with Powershell but that can lead to split brain

 

Clusters support automated failover, but then they have the safe guarding/witnesses involved etc.

 

Or you can use things like Veeam/other backup software to do the automation and witnessing.

 

Steve

[sonofsanta]

Are you talking about virtual hosts, here, not virtual guests?

 

If so, you don't have a backup server, as such - you have equal partners in a cluster.

 

We have two hosts with gobs of RAM & processor cores and each of them has, at any point, about half of the VMs. If one host goes down, the other host takes over the running of the VMs and they all run on that while you sort out your failed host. When you bring that host back up, nothing then happens until you tell it to - they'll continue to sit on the one host until you start live migrating them back over.

 

There's no worries about data being written to two places because it's not, it's written to shared storage -- in our case, a SAN. I think 2012R2 has all sorts of new funkiness to let you do this without a SAN, though I'm not up to speed on that at the moment (we're on 2008R2 until next year, when these servers turn 5 and get replaced).

[strawberry]

Your thinking of two separate things here.

failover - Shared storage, different hosts.

replication - Different storage, different hosts.

 

Failover will work in real time, one node fails the other starts the services.The Vm's can be on any node and split as you like. You can manually failover Vm's without loss of service to do maintenance etc.

Replication will only failover at your request, then you reverse the sync and start the original back up at your leisure.

[DCUK6]

Ok so that make sense.

 

Forgot about SANs as it not something we could afford for a while. If we did then you would store the VM on the SAN and if server 1 went down, server 2 would start up using the same virtual driver on the SAN as server 1.

 

Im thinking about something like two server with dual cpu and lots of ram and HDD space replicating the VM files to the HDD on both servers. So this would be manual startup. Does Hyper V do the replication or would you setup DFS-R?

 

 

Im asking like im about to do it but im just interested and wanted a techy chat as I miss that since moving to this place on my own .

May do it in the future though.

[Steve21]

You need some kind of shared storage for a cluster as mentioned as it's only one set of files that both hosts look at (and obviously having them stored locally on 1 HOST would be a failure point etc)

 

In your scenario best idea "imo" would be have a 50:50 split, so half on each, replicating to other. That way shares out resources of the HOSTs but gives the backup if needed. Then you'd manually failover the servers (using the failover option), so if the other comes back it knows etc.

 

DFS would only be useful in that sense if you have two of each server running at once, so for some things e.g. SIMs would be useless

 

Steve

[strawberry]

the shared storage need not be a san , you could use iscsi on a windows server to create a share. You could even use upwards of 1 server, ie have 2 or 3 shares across different servers in the same cluster. This is fairly cheap as the nodes could be cheapo ones from ICT direct with loads of processor and ram, and 3 cheapo disks and then buy a decent server with 8x146gb disks.

 

Alternatively, and I think this is the one, you could have two nodes, each with 50% of the vms on, and replicating to the other. It's not failover but in the event of one failing you could start the servers on the other node manually. the replication would be automatic, though as with anything else that's critical it would need checking once a day to ensure it is replicating.

 

Go get 3 decent desktops and go build it! .

[DCUK6]

Cheers,

 

Just dragging out some old laptops to give it a go.

 

Was going to build three virtual box 2012 machines but you cant have hyper v inside a virtual box machine. Surrounded by so many machines right now.

 

Our LEA might pay for our internet upgrade so now have 2.5k to use elsewhere. Thinking of a new server to put this into action.

Second option sounds more simple and not so reliant on a single file server. Does get rid of replication though.