Description
We see a lot of failures on Travis because of Timeouts.
I think the problem is simply that the default ask timeouts of "10 seconds" is too short to reliably execute tests in parallel on the small containers on Travis, especially when many services (like zookeeper mini clusters, etc) are involved. The tests work most of the time, but with the large amount of tests we have, builds fail if 1 out of 1000 tests experiences a timeout.
I suggest that we change the ForkableMiniCluster such that it multiplies the default timeout by a factor from an environment variable, which we set in the travis build scripts.
Something like export TEST_TIMEOUT_FACTOR=3.