Comment 1 for bug 1253896

logstash says this is a persistent flaky bug. We have a number of these, including the console log one, that all have one thing in common: they involve expected activity on the booted vm between when the vm is reported as 'active' and when it is "really booted". I checked in logstash and this never happens in neutron, except for the tests that have been trying to run neutron in parallel. So this is likely an issue of super-slow response from the vm both due to the load on the host and because the vm is nested. I saw that the banner checking code in paramiko allows 15s for the first line to appear but only 2 seconds (hard-wired) for each subsequent line. Also, the code in tempest that does the ssh connection catches socket errors (presumably for when the vm networking is not up yet) and AuthenticationException and then retries. But it does not catch SshException.

I will try catching this exception and see if that fixes the problem. The console log bugs might be similar in that they are demanding that 10 lines show up but perhaps sometimes that just takes a really long time. Please tell me if this makes no sense.