High Availability and Disaster Recovery (HADR) at the same time

When to combine high availability and disaster recovery (HADR)?

HADR architecture with a fast network

2 nodes are put in 2 remotes sites for implementing at the same time high availability and disaster recovery (HADR).

HADR solution with a fast nework

A solution without shared disk like Evidian SafeKit is required with 2 nodes replicating data synchronously and in real-time.

Because there is no data loss with synchronous replication, an automatic application failover can be implemented in case of failures.

Why an extended LAN/VLAN?

• Because there is an automatic failover of the virtual IP address with 2 nodes in the same subnet.
• Because a good latency (typically a round-trip of less than 2ms) is required for synchronous replication with no data loss.
• Because a good bandwidth (typically 1 Gb/s or more) is required for resynchronizing data at failback

When NOT to combine high availability and disaster recovery (HADR)?

HADR architecture with a slow network

2 nodes are put in the first site for high availability and a backup solution is implemented for disaster recovery.

HADR solution with a slow network

A high availability solution like Evidian SafeKit is implemented in the first site with synchronous real-time replication and automatic failover.

And a backup solution is implemented for asynchronous replication through the slow network to the disaster recovery site.

Failover to the disaster recovery site

• As there is data loss with a backup solution, the failover is manual and decided by an administrator.
• Backups are restored on servers at the disaster recovery site. Today, virtual machine backups are often implemented to facilitate this restoration (like Veeam).
• Rerouting to the disaster recovery site is made at the DNS level. The rerouting time depends on DNS cache timeout. And this may require restarting clients that perform their DNS resolution only once during their initialization.

High availability does not remove the need for a backup solution

Real-time replication of high availability solutions like SafeKit does not remove the need for a backup solution. Some events are recoverable only with a backup solution. For example, a ransomware encrypting replicated data on the primary server will encrypt data on the secondary with a real-time replication. Only a backup solution with a retention policy can solve a ransomware attack.