DanielLeens commented on issue #11583: URL: https://github.com/apache/seatunnel/issues/11583#issuecomment-5111113314
Thanks for writing this down with the failure boundary separated from the first causal exception. I checked the current E2E infrastructure, and this is a valid test-reliability problem: the shared Flink E2E containers still bind the JobManager UI as `8081:8081`, and many connector E2E modules have similar fixed host-port bindings. So a failed or slow cleanup in one invocation can indeed make a later invocation fail with `port is already allocated`, even when the fixed port was not the original root cause. PR #11574 looks like the right repair track for this issue. Because it touches a broad part of the E2E matrix, I think the review should focus on four boundaries: 1. container-to-container traffic should continue to use Docker network aliases and service ports; 2. clients running in the test JVM should use Testcontainers mapped host ports; 3. services with advertised addresses or clustered protocols, such as Redis Cluster, HBase, TiDB, and similar systems, need explicit endpoint rewriting or forwarding instead of a simple mechanical port removal; 4. the change should preserve the first failure signal rather than only making the later port-collision symptom disappear. The current #11574 head already has the right direction, but its `Build` check is failing, so the next gate is to fix that CI failure and then review the higher-risk clustered services carefully. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
