kestone bootstrap fails if previous keystone running

[Vladislav Belogrudov]

Sometimes keystone service is running on a bootstrapping host due to previous ansible attempts (failed/misconfigured or old service). Bootstrap container creates database but fails to create endpoints/users because it tries to bind temporary keystone to already bound port. Ansible should delete running keystone before bootstrapping if the last one is necessary

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/211158

[Steven Dake (sdake)]

I think the proper way to handle this scenario is through a new integration point rather then just deleting in the current playbooks. The proposed patch exposes the scenario where a rerun of the playbooks would result in lost data without the operator's knowledge. This would be highly surprising.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Vladislav Belogrudov (<email address hidden>) on branch: master
Review: https://review.openstack.org/211158
Reason: Too much control over failed task

[Vladislav Belogrudov (vlad-belogrudov)]

I ran kestone initially and it ran OK but did not create anything due to 'su' problem - some kernels don't allow 'su' inside of a container (also the same can happen during table migration and many other possibilities - lack of space, memory,...). To reinit the service I had to delete user/database and run ansible playbook again.

Status from bootstrapping container is not tracked so far so normal keystones started after the former exited. If status of bootstrapping could be taken into account the problem would not occur.

[Sam Yaple (s8m)]

I believe we discussed this on IRC at one point.

If the bootstrap process failed initially and the container never properly bootstrapped, it is reasonable to require the operator to remove that database (that is empty) before continuing. We do _not_ want to accidentally start bootstrapping again.

Since this is only an issue that only occurs during an initial bootstrap, I believe this is best left to documentation. I am open for other suggestions.

[Vladislav Belogrudov (vlad-belogrudov)]

agree, not a bug

[Michał Jastrzębski (inc007)]

I'm reoppening this bug as we will want to run bootstrap during upgrade, as we need to do db migrate prior to upgrade of containers

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/257567

[OpenStack Infra (hudson-openstack)]

Change abandoned by Michal Jastrzebski (inc0) (<email address hidden>) on branch: master
Review: https://review.openstack.org/257567
Reason: it's done in playbooks already

[Christian Berendt (berendt)]

Related review was abandoned because it is already done in the playbooks. Closing as invalid.