Geo-Replication and Failover Groups | Guided by A Guide to Cloud

Cross-region disaster recovery

Simple explanation

Active geo-replication is like having a live copy of your office in another city. Everything you write in your primary office is instantly copied to the remote office. If your primary burns down, the remote office has everything.

Failover groups take it further — they add a shared phone number (listener endpoint). Callers always reach the active office automatically. When you fail over, the phone number follows the active office.

Active geo-replication

Available for Azure SQL Database only (not MI, not VMs):

How it works

Primary database in Region A
Create up to 4 readable secondary databases in other regions
Asynchronous replication — RPO typically under 5 seconds
Secondaries are readable (offload read queries)
Failover is manual — you initiate it

# Create a geo-secondary
az sql db replica create \
  --resource-group NorthStarRG \
  --server northstar-sql-primary \
  --name NorthStarERP \
  --partner-server northstar-sql-secondary \
  --partner-resource-group NorthStarRG-DR

-- Check replication status
SELECT
    partner_server, partner_database, replication_state_desc,
    last_replication, replication_lag_sec
FROM sys.dm_geo_replication_link_status;

Failover

# Manual failover (secondary becomes primary)
az sql db replica set-primary \
  --resource-group NorthStarRG-DR \
  --server northstar-sql-secondary \
  --name NorthStarERP

Failover groups

Failover groups provide automatic failover and a listener endpoint that applications use — no connection string changes needed during failover.

Geo-replication vs failover groups

Geo-Replication vs Failover Groups
Feature	Active Geo-Replication	Failover Groups
Platform	SQL Database only	SQL Database and Managed Instance
Automatic failover	No (manual only)	Yes (configurable)
Listener endpoint	No	Yes — read-write and read-only endpoints
Max secondaries	4 per database	1 secondary region
Multiple databases	Individual database	Group of databases fail over together
Connection string	Must update on failover	No change — listener redirects
Grace period	N/A	Configurable (min 1 hour)
RPO	< 5 seconds	< 5 seconds

Setting up a failover group

# Create failover group for SQL Database
az sql failover-group create \
  --resource-group NorthStarRG \
  --server northstar-sql-primary \
  --name northstar-fog \
  --partner-server northstar-sql-secondary \
  --partner-resource-group NorthStarRG-DR \
  --failover-policy Automatic \
  --grace-period 1

# Add databases to the group
az sql failover-group update \
  --resource-group NorthStarRG \
  --server northstar-sql-primary \
  --name northstar-fog \
  --add-db NorthStarERP

Failover group endpoints

Endpoint	Pattern	Connects To
Read-write listener	`northstar-fog.database.windows.net`	Current primary
Read-only listener	`northstar-fog.secondary.database.windows.net`	Current secondary

Priya’s setup at ScaleWave: Application connection strings point to the failover group listener, not the individual server. When failover happens, the listener automatically routes to the new primary — zero application changes.

Failover policies

Policy	Behaviour
Automatic	Azure detects primary outage and fails over after grace period
Manual	You initiate failover explicitly
Grace period	Minimum time before automatic failover (1+ hours) — prevents false positives

Managed Instance failover groups

MI failover groups work similarly but with key differences:

Secondary MI must be in a different region
VNet peering or VPN required between MI subnets
All databases on the MI fail over together (instance-level, not per-database)
DNS alias handles connection redirection

Exam tip: failover group vs geo-replication

The exam loves this comparison. Quick decision:

Need automatic failover? → Failover group
Need listener endpoint (no connection string change)? → Failover group
Need 4 readable secondaries? → Geo-replication
Need MI DR? → Failover group (geo-replication doesn’t support MI)
Need per-database granularity? → Geo-replication (failover groups are all-or-nothing)

Question

What does a failover group listener endpoint provide?

Click or press Enter to reveal answer

Answer

A stable DNS name that always points to the current primary. Applications connect to the listener, and it automatically redirects to whichever server is currently primary — no connection string changes during failover.

Click to flip back

Question

Active geo-replication vs failover groups: which supports Managed Instance?

Click or press Enter to reveal answer

Answer

Failover groups support both SQL Database and Managed Instance. Active geo-replication is SQL Database only.

Click to flip back

Question

What is the grace period in a failover group?

Click or press Enter to reveal answer

Answer

The minimum time Azure waits before triggering automatic failover after detecting a primary outage. Default is 1 hour. Prevents false positives from transient network issues.

Click to flip back

Knowledge Check

Priya needs Azure SQL Database DR that fails over automatically with no application connection string changes. Which solution should she configure?

🎬 Video coming soon

Next up: Always On: Availability Groups and FCIs — configure Always On for SQL Managed Instance and Azure VMs.