The python tests don't retain the server logs. According to https://github.com/apache/kudu/blob/master/python/kudu/tests/common.py it looks like we are sending the logs to stderr but then we throw away the output. The output doesn't end up in the python3_client.log, even though build-and-test.sh has this:
This lack of logs makes it hard to debug python test failures, especially when they happen on pre-commit builds in Jenkins.
I also wonder if we are buffering stderr and if it's too large for the buffer then the daemon can block. There was a hang on one of the python tests and I don't know why; this is just speculation. It could have also been some kind of server or client deadlock.