Question 34 of 50
What is Retry?
Question
"What is Retry in the context of Kafka message processing?"
What the interviewer wants to assess
Whether you distinguish transient from permanent errors — the basis for knowing when retry helps and when it's a waste of attempts.
Resposta rápida
Retry is trying to process a failed message again, expecting the error to be transient (a network timeout, a momentarily unavailable database). Usually combined with backoff — a growing interval between attempts — to avoid overloading an already unstable system.
Resposta nível Sênior
Retry only makes sense for transient errors — failures with a real chance of not repeating on a new attempt. For permanent errors (a malformed payload, a business rule that always rejects that data), retry is useless: it will fail the same way indefinitely, unless there's a maximum attempt limit, after which the message moves to a DLQ. In Kafka, the most robust implementation uses a dedicated retry topic — the failing message is republished to it with a header indicating how many attempts have already happened, and a separate listener reprocesses it after the backoff, without blocking the main topic (avoiding the poison pill effect).
In-depth explanation
See "Retry and Backoff" and "Retry Topic" in Chapter 9.
Exemplo financeiro
If the saldo-service fails to process an event because the database is momentarily unavailable, Spring
Kafka retries with 1s, 5s, and 30s backoff before considering the message unrecoverable by simple retry and
routing it to the DLQ.
"Retry solves any kind of error, just try more times"
Retry only solves transient errors. For a permanent error (e.g., a payload that will never deserialize correctly), retrying indefinitely just delays detecting the real problem and can stall processing of the following messages.
Pode vir a seguir
Likely follow-ups: "what is DLQ?" and "what is a Poison Pill?".
Related chapters