AWS CLI - Looping through DMS Migration task ARNs to wait

[eternaltyro]

In Github Actions, I tried looping through a list of DMS Migration Tasks (ref by ARN) to check their status using aws dms wait. But I always get Waiter encountered a terminal failure state error.

Waiter ReplicationTaskStopped failed: Waiter encountered a terminal failure state: For expression "ReplicationTasks[].Status" we matched expected path: "starting" at least once
Error: Process completed with exit code 255.

The run step on Github Actions looks like this:

      - name: Trigger DMS Replication Tasks
        run: |
          # Define the list of DMS task ARNs
          TASK_ARNS=(
            "arn:aws:dms:eu-west-2:<AC_ID>:task:<ARN_1>"
            "arn:aws:dms:eu-west-2:<AC_ID>:task:<ARN_2>"
            "arn:aws:dms:eu-west-2:<AC_ID>:task:<ARN_3>"
            "arn:aws:dms:eu-west-2:<AC_ID>:task:<ARN_4>"
          )

          # SUCCEEDS
          for ARN in "${TASK_ARNS[@]}"; do
            echo "Starting DMS Task: $ARN"
            aws dms start-replication-task \
              --replication-task-arn "$ARN" \
              --start-replication-task-type reload-target
          done

          # FAILS
          for ARN in "${TASK_ARNS[@]}"; do
            echo "Checking DMS Migration Task Status: $ARN"
            aws dms wait replication-task-stopped \
              --filters Name=replication-task-arn,Values="$ARN"
          done

[uldyssian-sh]

aws dms wait replication-task-stopped is the wrong waiter for what you’re doing.

In DMS, start-replication-task transitions the task through states like starting and then into a running state (running / load-complete / etc.). The waiter replication-task-stopped is designed to succeed only when the task ends in a “stopped” state, and it treats other states (including starting) as terminal failure conditions. That is why you see:

Waiter ReplicationTaskStopped failed ... matched expected path: "starting"

If your goal is “wait until the task finishes the reload and is no longer running”, use the waiter that corresponds to the end state you expect, or poll for the exact desired status.

Two correct approaches:

Wait for the task to be RUNNING (if you just want to ensure it started):

aws dms wait replication-task-running
--filters Name=replication-task-arn,Values="$ARN"

Or, if you want to wait until the reload is completed, poll describe-replication-tasks for a completion status (example: load-complete for full load tasks):

STATUS=$(aws dms describe-replication-tasks
--filters Name=replication-task-arn,Values="$ARN"
--query "ReplicationTasks[0].Status"
--output text)

Then loop until STATUS is one of your expected “done” states (commonly load-complete, or stopped if you explicitly stop it), and treat failed as an error.

In short: replication-task-stopped will fail for a task that is legitimately in starting/running. Use replication-task-running to wait for startup, or poll for the completion state you actually want (e.g., load-complete) instead of waiting for stopped.

[uldyssian-sh]

Thanks! If this helps, please consider marking it as “Helpful” or “Accepted” 😊