1

We are planning to create a backup data center for warm disaster recovery. Critical services will be replicated on application level to the new data center, all other services will be backuped daily.

For external services, we will use an external failover provider. And I have no idea how to manage internal services.

For example, clients connect to app.company.com which is 10.0.0.10. In case of disaster, how would clients understand that they need to connect to 10.0.1.10?

2
  • 1) Won't FHRP be sufficient? It needs to monitor the availability of the DC through ping or L7 probes. 2) I'm also thinking of LISP if it may help. Eventually the main goal of LISP is to have the machine maintain the primary IP but be located in a different place. In the OP's case the physical location would be the same but from the topology point of view it will behind a different location, maybe just routing should be redefined Oct 18 '21 at 12:11
  • Did any answer help you? if so, you should accept the answer so that the question does not keep popping up forever, looking for an answer. Alternatively, you could post and accept your own answer.
    – Ron Maupin
    Dec 23 '21 at 21:31
0

Using DNS names and changing those on the fly leaves quite a time gap due to DNS caching.

On the (on-topic) network/transport layers, you could either use NAT across a load balancer (cluster) and hide the real host addresses behind that (easy but awkward due to NAT, might require some beefy hardware), or use virtual IP addresses (much more elegant):

  • Each server uses two addresses: a dedicated one for management & synchronization, and a virtual one, shared with its failover partner and used by clients
  • For a failover, you simply change the routing of the virtual address range (or for single addresses). That could be done by route advertisement or on an in-front router (cluster).

In any case, make sure you have a good recovery scheme to return to normal operation - after a failover, the synchronization / replication direction needs to be reversed temporarily to update the main production servers. This is often much more difficult than mirroring application data across locations.

0

If your users surf through a firewall, they sometimes have the ability of load balancing etc. Speaking of FortiGate devices, there it's called Virtual Server.

You give it an IP address, the one the DNS server points to and then you give it one or more IP addresses, those of the servers which host the service. You can select different methods of load balancing and even do health checks for those IPs.

The downside is, that (AFAIK) the traffic goes through this firewall in any case: Let's say you have SiteA and SiteB, your main service and Firewall are located in Site1, your load balancer points your traffic to an IP in SiteA and to the secondary IP in SiteB. Now if not only the service fails but the whole uplink or infrastructure of SiteA, and the Firewall of SiteA won't be available, than the load balancer won't direct traffic to SiteB obviously.

3
  • The question is about warm failover across data centers. That includes the gateways.
    – Zac67
    Sep 17 '21 at 14:42
  • Oh, you're right. Nevertheless, some companies have a central firewall connecting its data centers, hosted even by the provider and often referred to as VDOM, which definitely might have this type of capabilities. In this case it would be ideal.
    – josibu
    Sep 17 '21 at 15:02
  • Sure, but that wasn't really the question. Btw, on a Fortigate a VDOM is a virtual management and routing domain that may be delegated to other administrators. Central firewalls are usually dedicated.
    – Zac67
    Sep 17 '21 at 15:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.