Conversation
Partition consumer goroutines were spawned before k.msgChan was assigned, causing a data race between the goroutine reads (line 104) and the subsequent write (line 304). Move the assignment before the spawn loop. Fixes CON-407
TestRedpandaRecordOrderSoakTest had a hardcoded 3-minute soak duration but didn't account for time already consumed by other tests in the package. When run with a 5-minute timeout alongside other kafka integration tests, the test was killed before it could finish. The soak duration now adapts to the remaining time budget from t.Deadline(), reserving 2 minutes for container setup/teardown. If insufficient time remains (<30s for soaking), the test skips with a clear message. Fixes CON-413
KafkaSeedBroker() returns localhost:<port> which fails when the system DNS resolver cannot resolve "localhost" (e.g. under Docker networking issues). Use 127.0.0.1 to bypass DNS resolution entirely. Fixes CON-425
range_of_partitions was nested inside explicit_partitions as a child t.Run. Go's test runner releases PAUSEd subtests only when the parent function returns, so the direct explicit_partitions subtests were blocked until range_of_partitions completed (~19s), leaving them with a tight remaining deadline that caused flaky 10s timeouts. Moving range_of_partitions to a sibling t.Run ensures both groups get their own share of the parent deadline. Fixes CON-426
TestRedpandaIntegration runs 5 sequential suite groups that each include 1000-message sequential and parallel tests. Later groups (especially manual_partitioner) start with a nearly exhausted parent test deadline because preceding groups consume most of the available time. Partition-specific groups only need to validate routing correctness, not throughput. Reduce their message count from 1000 to 100, keeping the top-level suite at 1000 for thorough coverage. Fixes CON-427
Lower batches from 1M to 100K (still 1M messages) so the test completes in ~2s instead of ~20s, avoiding timeout when CPU-starved by long-running integration tests in the same package. Fixes CON-431
TestIntegrationUnordered runs 4 sequential suite groups that each include 1000-message sequential and parallel tests. Later groups (especially manual_partitioner) start with a nearly exhausted parent test deadline because preceding groups consume most of the available time, causing the redpanda container to become unresponsive under sustained load. Partition-specific groups only need to validate routing correctness, not throughput. Reduce their message count from 1000 to 100, keeping the top-level suite at 1000 for thorough coverage. Mirrors the same fix applied to TestRedpandaIntegration in edb48668f. Fixes CON-439
TestIntegrationUnordered's manual_partitioner group did not call t.Parallel() like its sibling subtests, so it ran sequentially after the top-level suite, adding to the total elapsed time of the kafka integration package. The go test binary-level 5-minute timeout would fire before TestIntegrationUnordered could even complete its pre-test topic creation. Call t.Parallel() in manual_partitioner to match the other groups, and bump the kafka package timeout to 10m in packages.json, matching other heavy integration packages (redpanda, mssqlserver, etc.). Fixes CON-443
TestIntegrationUnorderedSasl's stream/parallel groups each ran 1000 messages, the same as the non-SASL TestIntegrationUnordered. The SASL variant only validates authentication wiring, not throughput, so the heavy message counts just consumed the parent package budget with no additional coverage. When other tests in the kafka package run first under the package timeout, the SASL subtests were left with no time to even connect to the broker. Drop the message count from 1000 to 100 to match the pattern used by the partition-specific groups in TestRedpandaIntegration (commit edb48668f). Standalone runtime drops from ~18s to ~6s. Fixes CON-440
TestRedpandaRecordOrderSoakTest used to consume ~3 minutes of soak (plus 2 minutes of setup/teardown budget) per run, dominating the kafka package runtime. When it ran before TestRedpandaRecordOrderIntegration under the package timeout, the latter was starved and killed by the package-level test timeout before it could complete. The chaos scenarios (leadership transfer + broker restarts) produce ample coverage within 1 minute, and the comment at the top of the test already notes that long soaks are meant to be run explicitly via `-count 1000` overnight. Standalone runtime drops from ~184s to ~72s. Fixes CON-448
|
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.
Commits
Jira