Conversation
ilidemi
commented
Apr 15, 2026
- Cleanup of a toxi test to avoid port conflict on the next run
- Limit PG slot names as they silently get truncated and chop off the suffix, enforce that for future tests
- QoL improvement to not rebuild flow services on test file changes
❌ 2 Tests Failed:
View the top 2 failed test(s) by shortest run time
To view more test analytics, go to the Test Analytics Dashboard |
❌ Test FailureAnalysis: Three e2e tests deterministically fail because the generated PostgreSQL replication slot names exceed the 64-character limit (65–66 chars), which is a real bug in slot name generation for tests with long names. |
❌ Test FailureAnalysis: BQ tests consistently fail because generated PostgreSQL replication slot names exceed the 64-character limit (caught by new validation in this PR), and TestCancelTableAdditionRemoveAddRemove fails because column c3 is missing from the destination after a cancel/re-add sequence — both are deterministic real failures, not timing-related flakiness. |
🔄 Flaky Test DetectedAnalysis: The e2e ClickHouse test failed with "SetupCDCFlowStatusQuery stuck in snapshot somehow" after 737 seconds, indicating a timing/race condition where the CDC flow failed to transition from snapshot to streaming mode — a known flaky pattern in Temporal-based integration tests. ✅ Automatically retrying the workflow |
❌ Test FailureAnalysis: The test |
🔄 Flaky Test DetectedAnalysis: The test ✅ Automatically retrying the workflow |