redpanda: report per-record errors from output batch writes#4400
Open
twmb wants to merge 1 commit into
Open
Conversation
The franz-go-based redpanda output's WriteBatch returned a single joined error whenever any record in the batch failed, which forced Connect to re-deliver the whole batch and duplicate every record the broker had already acknowledged. Switch to service.NewBatchError(...).Failed(idx, err) -- the same pattern used by the legacy sarama kafka output -- so Connect can re-deliver only the actually-failed records. This makes the auto_replay_nacks: false + fallback DLQ pattern usable for partial batch failures. Also fix a concurrent append data race on the shared errs slice by indexing into a preallocated per-record error slice from each Produce callback. Refs #4387. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
Commits Review LGTM |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
redpandaoutput'sWriteBatchreturned a single joined error whenever any record in the batch failed, forcing Connect to re-deliver the whole batch — which duplicates every record the broker had already acked. Switch toservice.NewBatchError(...).Failed(idx, err), same pattern as the legacy saramakafkaoutput (internal/impl/kafka/output_sarama_kafka.go:619), so Connect can re-deliver only the actually-failed records.auto_replay_nacks: false+fallbackDLQ pattern (the supported strict-ordering recipe documented on theredpandainput) actually usable for partial batch failures: the DLQ now receives only the failed records instead of duplicates of every record in a partially-failed batch.append(errs, err)from multiple Produce callbacks concurrently. The new code indexes into a preallocated per-record[]errorslot, so each callback writes to its own index.Refs #4387.
Test plan
go test -count=1 -race -run TestBatchWriterPartialFailures ./internal/impl/kafka/go test -count=1 -short ./internal/impl/kafka/...gofumptcleango vet ./internal/impl/kafka/...The new
TestBatchWriterPartialFailuresis a table-driven unit test exercising:SkipRecordsentinel → returnsnil.DecorateRecord→BatchErrorwith all indices.DecorateRecordfailures +SkipRecordskips →BatchErrorwith only the decorate-fail indices (skips do not count as errors).kgo.Client) →BatchErrorwith all indices.BatchErrorwith only the produce-fail indices.BatchErrorwith both sets of indices.