Failover between 2 datacenters?

[kb1]

Hello Guys,

Is it possible to configure failover between two datacenters , right now we have one cluster in one data center(which are our internet firewalls) and planning to add another cluster in the second datacenter, now how do we achieve a failover? Can it be achieved if we add the second cluster in the second datacenter to the existing cluster? And if not what solution would work for a proper failover between 2 datacenters?

Thank you.

[PhoneBoy]

Failing over between sites almost always means dynamic routing is involved.
Also, unless you are using provider-independent IP space, you generally cannot do any sort of failover gracefully as the IPs will have to change when you switch sites.
Even then, you may still have issues as BGP updates on the Internet.
None of this involves clustering the gateways at each site together.
Also, clusters have to share the same L2 networks, which they often don't in a cross-site failure.
However, we do support cross-site clustering with different L2 domains in R80.40.
It doesn't solve any of the above issues, though.

However, if you do want active/active clustering that works across sites, you need to be in R80.40

[kb1]

Hi if you look at pg 36 of cluster_xl admin guide for R80.10 (we are using R80.20 by the way) you can see some points related to distribution of clusters over a wide geographical network (which is what it would be for clusters between 2 datacenters) but I do not see any mention related to bgp issues, etc or the issues that you are referring to(it does mention that only layer 2 should be involved between cluster interfaces), or maybe I'm not reading properly? Maybe it's mentioned somewhere else? Please do tell, if it's not a viable solution then we will not go for it.

Regards.

 

[PhoneBoy]

Cluster members have to have Layer 2 connectivity on all interfaces prior to R80.40.
That implies the same subnets are available at each site as well, including external networks.
For external-facing addresses, this usually means provider-independent IP addresses.
The routing of that between sites has to be changed through dynamic routing.
Likewise, depending on how your internal network is set up, you may need dynamic routing for a failover to occur cross-site as well.

Bottom line: you need to understand how the rest of your network can react to this.
Cross-site clusters can be helpful (subject to limitations) but it requires things elsewhere in the network as well.

[kb1]

Thank you

[kb1]

However I do get your point regarding bgp issues, changing of ips, etc.