… and busy responses
- Add HTTP workers section describing the deferred ingress queue
- Document which log processors run inside HTTP workers and the fallback
- Document 503 / gRPC status 14 responses when the ingress queue is full
- clarify retry guidance for ingress queue rejections
- Explain that mem_buf_limit can lower the effective ingress queue byte limit
- Separate the temporary queue-full rejection from the oversized-request rejection
- Recommend retrying only when the queue is full; oversized requests need smaller requests or a higher limit
- Compare the effective byte limit against each queued event instead of the whole request
- Note that one request can add several events to the ingress queue
- Align the retry guidance with the per-event limit
- Note that clients can't distinguish a full queue from an oversized event in the response
- Recommend retrying with backoff and explain that repeated failures point to an event exceeding the effective byte limit
Signed-off-by: Eric D. Schabell <eric@schabell.org>
Signed-off-by: Eric D. Schabell eric@schabell.org