Recent builds have failed due to a failure in test_breakpad.py. Assigning to Tim as the person who most recently touched this file.
Distilled TEST-impala-custom-cluster.xml output:
Notice that the main thread appaars to be killing statestore, but fails to kill impalad. Notice that a message appears that says that all impalads are running in the midst of the code that tries to shut down the cluster. Is this test multi-threaded? Is there more than one “main thread” Are these main threads working at cross purposes? What recent change may have caused this?
Also, looks like the script is sending signal 10 (SIGUSR1) while the statestore (in its log) says it got a SIGTERM (15):
Not terribly familiar with this area of the product, so bumping it over to the BE team.