View a markdown version of this page

Running a recovery plan - AWS Elastic Disaster Recovery

Running a recovery plan

You can run a plan as a drill or as a recovery:

Drill

Launches drill instances to validate your recovery readiness. Drills do not affect your source servers or ongoing replication. Run drills regularly so that your plan is proven before you need it.

Recovery

Launches recovery instances for an actual recovery event.

A plan recovers each server the same way an individual recovery does. Any post-launch actions that are active for a source server run for that server's recovery instance exactly as they would for an individual recovery. For more information, see Post-launch action settings.

Important

AWS Elastic Disaster Recovery does not terminate the instances that a plan launches, including drill instances, and it does not clean them up when an execution completes, is canceled, or times out. Every instance a plan launches keeps incurring Amazon EC2, Amazon EBS, and associated charges until you terminate it. Because a plan can launch up to 100 instances at once, clean up after every drill. You can terminate recovery instances from the Recovery instances page; see Managing recovery instances and Post recovery drill actions. For charge details, see AWS Elastic Disaster Recovery pricing.

To run a recovery plan (console)

  1. Open the Elastic Disaster Recovery console at https://console.aws.amazon.com/drs/home.

  2. In the navigation pane, choose Recovery plans, and then choose the plan that you want to run.

  3. Choose Execute recovery plan.

  4. Choose Drill or Recovery.

  5. (Optional) Choose a specific recovery point for individual servers. If you do not choose a recovery point for a server, AWS Elastic Disaster Recovery recovers that server from the latest data available. That is the same behavior as recovering the server on its own without choosing a recovery point.

  6. Confirm your choices to start the execution.

To start an execution with the AWS CLI, use the start-recovery-plan-execution command.

aws drs start-recovery-plan-execution \
    --recovery-plan-arn PLAN_ARN \
    --mode DRILL

To recover specific servers from a specific recovery point, add the --source-servers parameter. Every server that you list must belong to the plan and must include a recoverySnapshotID. You do not have to list every server in the plan. Any server that you omit is recovered from the latest data available.

aws drs start-recovery-plan-execution \
    --recovery-plan-arn PLAN_ARN \
    --mode RECOVERY \
    --source-servers '[
        {"sourceServerID": "s-EXAMPLE1", "recoverySnapshotID": "pit-EXAMPLE1"}
    ]'

A recoverySnapshotID is an AWS Elastic Disaster Recovery recovery point ID. It is always 21 characters long and begins with the pit- prefix, for example pit-1234567890abcdef1.

Important

A recovery point ID is not an Amazon EBS snapshot ID. An Amazon EBS snapshot ID begins with snap-, and although a recovery point lists the Amazon EBS snapshots that back it in its ebsSnapshots field, those snap- IDs are not valid values for recoverySnapshotID. Passing one is rejected as a validation error.

To find the recovery point IDs that are available for a source server, use describe-recovery-snapshots. Ordering by DESC returns the most recent recovery point first.

aws drs describe-recovery-snapshots \
    --source-server-id s-EXAMPLE1 \
    --order DESC \
    --max-results 5

Each entry in the response items list contains the recovery point ID in the snapshotID field, along with its timestamp. Take the value of snapshotID and pass it as recoverySnapshotID when you start the execution; the field is named differently in the two APIs.

{
    "items": [
        {
            "snapshotID": "pit-1234567890abcdef1",
            "sourceServerID": "s-1234567890abcdef1",
            "expectedTimestamp": "2026-08-18T09:00:00Z",
            "timestamp": "2026-08-18T09:00:12Z",
            "ebsSnapshots": [
                "snap-0123456789abcdef0",
                "snap-0123456789abcdef1"
            ]
        }
    ]
}

In this example, pit-1234567890abcdef1 is the value to pass as recoverySnapshotID. The two snap- IDs are the Amazon EBS snapshots that back that recovery point, and are not valid values.

To narrow the results to a time range, add the --filters parameter with fromDateTime and toDateTime.

Important

A recovery point that you pin is validated again at the moment its step begins, not only when the execution starts. In a long plan, a step can run many hours after you started the execution. A recovery point that existed then might be deleted, or age out of your point-in-time retention window, before the step runs. If that happens, the step fails and the plan stops. Pin recovery points only when you need a specific point in time, and prefer recovery points that will still be inside your retention window when the step is expected to run. Omit --source-servers to always use the latest data available.

The following conditions prevent an execution from starting. Each one returns the error shown:

Condition Error

The plan already has an execution in progress. A plan can have only one execution running at a time.

ConflictException — Another execution is already active for plan planId

A source server in this plan also belongs to another plan that has an execution in progress.

ConflictException — names the plan and the execution that are holding the server, so that you can wait for that execution or cancel it.

The plan status is not ACTIVE. A plan becomes INVALID when it has no server steps.

ConflictException — Recovery plan planId is in status status and cannot be executed. Only ACTIVE plans can be executed.

The plan has no steps.

ValidationException — Recovery plan planId has no steps. Add at least one step to the plan before starting an execution.

The plan is being deleted.

ConflictException — Cannot start execution: recovery plan planId is being deleted.

A server listed in --source-servers is not part of the plan, or is missing a recoverySnapshotID.

ValidationException — Source server IDs not found in plan or Missing recoverySnapshotID for servers, listing the servers concerned.

The AWS account has not been initialized for AWS Elastic Disaster Recovery.

UninitializedAccountException