backoff
Waiting longer and longer between successive retries, often doubling the delay each time, so a struggling service has room to recover.
In more detail
Backoff does not guarantee success; it stops one failure turning into a storm. Combine it with jitter, a maximum number of attempts, and any Retry-After the server sends.
Examples
The kubelet doubles the delay between container restarts up to five minutes; that is the BackOff in CrashLoopBackOff.A Kubernetes Job retries failed pods with backoff until backoffLimit.delay = min(maxMs, baseMs * 2 ** attempt)Related terms
- retry — Sending a request again after it failed or timed out, in the hope that it succeeds the second time.
- jitter — A random amount added to or taken from a retry delay so that many clients do not all retry at the same moment.
- Retry-After — The response header that tells a client how long to wait, in seconds or as a date, before trying again.
- thundering herd — A flood of requests that arrives at once when many clients react to the same event, such as all retrying after an outage.
- CrashLoopBackOff — A container status meaning the container keeps exiting after it starts, so the kubelet is waiting longer and longer before restarting it.
- ImagePullBackOff — A container status meaning the node cannot download the container image, so the kubelet is waiting longer and longer between attempts.
- Kubernetes Job — The Kubernetes object that runs pods until a set number of them finish successfully, for work that has a clear end.
- rate limiting — Capping how many requests a client may make in a period of time, and rejecting or delaying the rest.
- 429 Too Many Requests — The HTTP status meaning the caller has sent too many requests in a given time and should slow down.
Taught in
- Error Handling And Safe Retries — Modern API Architecture: REST, GraphQL & gRPC
- Rate Limiting Strategies — Modern API Architecture: REST, GraphQL & gRPC
- Webhooks Versus Polling — Modern API Architecture: REST, GraphQL & gRPC
- WebSockets — Modern API Architecture: REST, GraphQL & gRPC
- An Hour Without Payments: Continuously Reconciled — Certified GitOps Associate (CGOA) Exam Prep
- Choose The Right Workload Resource — Certified Kubernetes Application Developer (CKAD) Exam Prep
- Server-Sent Events: Streaming Updates & Recovery — Production Full-Stack Engineering
- Bidirectional WebSockets: Reconnection & Heartbeats — Production Full-Stack Engineering
- Fundamentals Sprint: Four Tickets Before The Doors Open — Kubernetes and Cloud Native Associate (KCNA) Exam Prep
- The Status Tells You Where To Look: Troubleshooting Pods — Kubernetes and Cloud Native Associate (KCNA) Exam Prep
- Jobs And CronJobs — Kubernetes: Production-Grade Container Orchestration
- Troubleshooting: Probes, Logs, exec And CrashLoopBackOff — Kubernetes: Production-Grade Container Orchestration
- Troubleshooting: A Timed Service Recovery Drill — Kubernetes: Production-Grade Container Orchestration
- Rate Limiting And Resource Consumption — Application Security & Threat Modeling for Engineers