A retry repeats an operation after an earlier attempt fails or its destination is unavailable. It helps systems tolerate fallible networks and transient connection problems. section 5.2.1 section 8.9.2
Fluent Bit output retries follow a scheduler-controlled process:
- Output plugins report OK, error, or a request to schedule a retry. section 5.2.1
- scheduler.base sets the minimum delay before another attempt; scheduler.cap limits the maximum delay. section 5.2.1
- The delay is randomized and expands exponentially, with an upper bound of min(scheduler.base * 2ⁿ, scheduler.cap). section 5.2.1
Configure HTTP-forward retries
Set scheduler.base to 2 seconds, scheduler.cap to 15 seconds, and retry_limit to 6. With the destination unavailable, the console reports repeated attempts and eventually fails.
Applications can also catch intermittent errors such as network timeouts. Suppress alerts below an error-rate threshold, but intervene when repeated failures indicate possible backpressure. section 8.9.2
