Zuul reports standard output/error still open after child exited

[Steven Dake]

The gating job reports the summary on an OSIC cloud provider. Here are the logs:

http://logs.openstack.org/37/412237/39/check/gate-kolla-kubernetes-deploy-centos-binary-2-ceph-multi-nv/a7ed001/console.html#_2017-01-16_15_35_14_981456

[Steven Dake (sdake)]

Clark Boyan did most of the debugging here:
http://eavesdrop.openstack.org/irclogs/%23openstack-infra/%23openstack-infra.2017-01-16.log.html#t2017-01-16T15:52:17

The solution is likely to rework the fd manipulation to match what is used in kolla-ansible since that seems simpler and more intuitive.

[Steven Dake (sdake)]

A secondary problem is that wait_for_vm in this particular run didn't report any type of error for visual inspection (other then 255). wait_for_vm failed, however, there is no visual indication as to what failed. It would require digging through the logs in more detail which is fine, but a pointer to that in the main gating scripts would be helpful :)

[Kevin Fox (kevpn)]

Redefining this to low priority, as nothing breaks when it prints this. So, its cosmetic.

If I had to guess, I'd say its probably because the helm server is sill running at the end of the run.

[Kevin Fox (kevpn)]

Actually, I think there is a fix for this in this PS:
https://review.openstack.org/#/c/413193/29

In the reboot job, it blocks the master's ssh connection from returning causing the job to fail. When I added a bit of scripting to kill off unexited processes, it continued ok.