Cluster Fault Tolerance Between Data Centres

P

parityboy

Limp Gawd
Joined
Nov 13, 2010
Messages
390
Hello --

I'm currently scoping a project for a customer whereby they want a pair of application servers in a fault tolerant pair behind a load balancer. No problem. However, they also want failover to another clustered pair in another data centre.

Accommodating one load balancer falling over is easy enough, however the application servers themselves sync via Apache Tribes, so the challenge (which I've never faced before) is how to keep them synced with the other pair. I'm wondering if a virtual LAN between the pairs (and "behind" the load balancers) would be the simplest approach?

I'd greatly appreciate any feedback from those who have experience of this. Many thanks. :)
 
Do they want an automatic failover or a "One-Button" type solution?

In our case we have a second DC as a DR site which veeam replicates all VM's to the second site, we can then do a one click failover which fails the vms to the dr environment and boots the machines up.
 
@k1pp3r

Thanks for the reply. To be honest, this was something that was mentioned in a very high level meeting with the S.I.; unfortunately the Managed Services bods weren't in the room. I know they use VMware as the virtualisation infrastructure, so I can only assume that Veeam is something they would also use.

However, the app servers are basically silos with no shared content between them, so I'd like to cover the angle of the cluster in the DR site as being a "warm spare" - fully in sync with primary cluster, but mostly idle. That way, if they have to cut over, the DR cluster is already up and running and synced with the primary.

Rather than syncing at the VM level, I'd like to sync at the app level if possible. Apache Tribes is LAN-only (for node discovery) as far as I am aware, so I'll probably need all four app servers (or at least one from each pair) to be able to contact each other "behind the scenes" without hitting the public-facing load balancer at each DC.

Curious: can Veeam do block-level updates from an in-use VMDK to a cold one? I'm thinking that if the solution we go for is similar to the one you mentioned, it would be nice to keep the DR site images as up-to-date as possible.
 
Veeam does its replication on the vmdk level to a cold virtual machine, in the standby environment you will see VM_Replica which is the replicated VM and it can replicate i think every 15 minutes, I only do 1 replication per day right now.

As far as app level, your gonna have to look at something like Arcserv or doubletake which are expensive as hell
 
how "in sync" do they require?
Do you need real time, or is a 15 minute sync interval sufficient?
If you can absorb a few minutes then the veeam solution might be the easiest. Plus it can take care of other VM's as well for backup and DR killing two birds with 1 stone.
 
cyr0n_k0r said:
how "in sync" do they require?
Do you need real time, or is a 15 minute sync interval sufficient?
If you can absorb a few minutes then the veeam solution might be the easiest. Plus it can take care of other VM's as well for backup and DR killing two birds with 1 stone.
Click to expand...

15 minutes is probably short enough, so the Veeam solution is viable. :)
 
You must log in or register to reply here.
Back
Top