13. High Availability and Disaster Recovery Concepts

Always On Failover Clustering

Up to 64 servers. Always On Failover Clustered Instance. FCI. Single site. Shared storage. One server owns the FCI. Active node.

Windows Storage Replica (2016+) geoclusters

VIP is assigned to active instance. Application servers connect to VIP

Quorum may include a witness device

Authorative partition /fq

Quorum models:

Can change Node Weight to zero

Dynamic Quorum - Tie Breaker

Always On Availability Groups (AOAG)

AOAG = Database mirroring plus cluster - stand alone instances

Needs availability group listener installed on the cluster

AOAG - best for small database with low write profiles - allows for synchronous commits on all nodes (HA). Up to 8 replicas implies a 9 node cluster including 3 synchronous replicas

AOAG (DR) - asynchronous mode

Failover at level of AG. Multiple AG per instance. Maximum (100 database amd 10 AG) per instance (Microsoft recommendation)

One database mirroring and point per instance - all log streams go through this i/p

Each AG has its own network name and IP address

Application reconnects after failover

Automatic Page Repair

Good page gotten from a replica except - file header, database boot, allocation

Log Shipping

Backup transaction log on primary, ship to secondary; restore it

Recovery Modes

Remote Monitor Server

Failover - backup tail log; copy remaining backup; apply them in order; last one with RECOVERY

Can combine AOFOC with AOAG

Synchronous mode allows automatic failover. Asynchronous mode only allows manual failover.

Can combine AOFOC with log shipping