test runner fails to reach the test bed

[Vincent Ladeuil]

Reported by Urshina on IRC:

INFO:tstrun.testbed:Got IP 10.0.0.109 for dec8d2e9-00de-40d5-9c5e-530110ca34d4
INFO:tstrun.testbed:cloud-init completed for dec8d2e9-00de-40d5-9c5e-530110ca34d4
INFO:image_test:Saving testbed console
INFO:image_test:The test bed is ready
INFO:image_test:Testing hello
ssh: connect to host 10.0.0.109 port 22: No route to host
ssh: connect to host 10.0.0.109 port 22: No route to host
INFO:image_test:Saving subunit results for hello
ssh: connect to host 10.0.0.109 port 22: No route to host
INFO:image_test:Saving adt-run log for hello
ssh: connect to host 10.0.0.109 port 22: No route to host
INFO:image_test:Saving adt-run summary for hello
INFO:image_test:Test completed for ticket 3

This appears to be caused by some glitch on hp cloud but may also be a race around the ssh server in the tesbed starting *after* cloud-init has completed.

In any case, the test runner should report a better error than empty summary, subunit and log files.

If the ssh server is slow to start, the tr worker should retry to connect before reporting an error.

[Vincent Ladeuil (vila)]

FTR, I'm ruling out the race between cloud-init and the ssh server. I was able to connect to a testbed long before cloud-init finished apt-get update/install. Given the other network related issues in the past with hp, I think it's safe to blame that instead.