DP-750 · 4.2.5 · Deploy and maintain data pipelines and workloads
Automatic restarts
Configure automatic restarts for a job or a data pipeline
Use task retries for transient failures and continuous execution when an always-on workload should restart after completion or failure.
See the job graph you are recovering
Automatic restart settings protect a job graph like this when infrastructure or transient failures interrupt execution. The diagrams below explain retry scope, restart behavior, and idempotency concerns.
Choose the smallest recovery scope
- One task has a transient failureTask retry
- Multi-task run partly failedRepair run
- Streaming job should keep runningContinuous mode
- Streaming query restartsResume from checkpoint
Retry only when another attempt has a realistic chance of succeeding.
Continuous mode restarts the job
- 1Run startsWorkload executes
- 2Run completes or failsScheduler evaluates state
- 3FailureBackoff increases after repeated failures
- 4New runJob starts again
Continuous mode uses job-level restart behavior and exponential backoff for repeated failures.
Preserve streaming progress
Structured Streaming checkpoints persist offsets and state so the restarted query can continue from prior progress rather than starting from zero.
Key points
- Task retry: Re-executes a failed task within policy limits
- Continuous job: Starts another run after completion or failure
- Backoff: Consecutive continuous failures use increasing delay
- Checkpoint: Lets streaming work resume from prior progress
Exam trap
Automatic restart is not a substitute for fixing deterministic failures that repeat every run.
Sources · 3 Microsoft Learn pages Show ▾
Checked on October 4, 2026. Report a problem
Go deeper on Microsoft Learn
Checked against Microsoft Learn on October 4, 2026. Report a problem with this card