transient failure
A failure caused by a temporary condition, such as a brief network glitch, that may succeed if simply tried again.
In more detail
Transient failures are the only ones worth retrying. A missing record or invalid input is not transient, and retrying it only adds load.
Related terms
- retry — Sending a request again after it failed or timed out, in the hope that it succeeds the second time.
- exponential backoff — Waiting progressively longer between retries, for example 1, 2, then 4 seconds, so a struggling service gets room to recover.
- dead letter queue — A separate queue where messages that keep failing are moved, so they stop blocking others and can be inspected and replayed later.
Taught in
- The Cost Of Checked Exceptions — Modern Java: Virtual Threads & High-Throughput Services
- Retries With Backoff Topics — Apache Kafka & Event Streaming