[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user kishorvpatil commented on the issue: https://github.com/apache/storm/pull/2660 @raghavgautam Can you please address the merge conflicts? ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user ghajos commented on the issue: https://github.com/apache/storm/pull/2660 @raghavgautam I completely missed something with my test. Originally started a tcp server and a tcp client in separate processes and sent KILLTERM to the client side. The server port stayed in TIME_WAIT then started a new client to check if it can connect to the server side. When #STORM-3039 is landed on our lines and the original exception was still present, rechecked the test and realized that even without the change the new client could connect to the server at least on ubuntu16. Sorry for the mistake! ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user raghavgautam commented on the issue: https://github.com/apache/storm/pull/2660 @ghajos Ping. ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user raghavgautam commented on the issue: https://github.com/apache/storm/pull/2660 @ghajos Can you tell me how you tested your patch ? ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user HeartSaVioR commented on the issue: https://github.com/apache/storm/pull/2660 @ghajos Thanks for the information. I didn't recognize the fact that the new test is passing without #2643. I agree this is out of scope, just think that if the new test can verify the new behavior "consistently", it is no reason to deny the new test. @raghavgautam Could you fix this, or just would like to simply close both the PR and issue? ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user ghajos commented on the issue: https://github.com/apache/storm/pull/2660 @raghavgautam Thanks for the quick reaction! Unfortunately the newly introduced test passes without #2643, so it does not test the original issue. Since in this modification both the server and the client is closed gracefully the port won't stuck in TIME_WAIT state, no need to use SO_REUSEADDR to restart new instances. To test #2643 the server port should stuck in TIME_WAIT state and then a new server instance should be started successfully. I still think that it is out of scope to test an operating system level TCP setting in a Storm test. cc: @HeartSaVioR ---
[GitHub] storm issue #2660: STORM-3056: Add a test for quickly rebinding to a port
Github user raghavgautam commented on the issue: https://github.com/apache/storm/pull/2660 @HeartSaVioR @ghajos Please review. ---
