Failover

Failover promotes the target site for one or more workloads in a protection plan. Use planned failover for maintenance windows. Use unplanned failover when the source site is degraded.

Pre-flight checks

Planned failover

1. Quiesce the source

   Stop application writes at the source if you require zero data loss. The platform finishes the in-flight replication cycle and emits a final recovery point.

   bash

   dm plan quiesce --name prod-tier1

2. Run failover

   Pick a recovery point and start the failover. The CLI prints the runtime job id.

   bash

   dm plan failover \
     --name prod-tier1 \
     --recovery-point latest \
     --network recovery-net

3. Validate at the target

   Run the validation suite that ships with the plan, then route real traffic.

   bash

   dm plan validate --name prod-tier1

Unplanned failover

When the source is unreachable, skip the quiesce step and run the failover against the latest available recovery point:

$ dm plan failover --name prod-tier1 --recovery-point latest --skip-source-quiesce

The platform fails the source quiesce step gracefully and proceeds. You will see a warning in the job output.

Recovery time

For a single Tier 1 VM with a warm recovery network, expect $T_{\text{recovery}}\le 10\text{minutes}$. For groups of dependent VMs, sum the longest path through the dependency graph.

After failover