python -m enos.enos destroy doesn't destroy the containers created during the upphase

[msimonin]

No description provided.

[rcherrueau]

The destroy only works if the up phase succeeds to complete. Reason is simple: destroy uses environment information to destroy resources and environment information are written at the end of the up. If the up phase fails, then there is no environment information.

I could maybe fix that by passing an extra argument to each phase. This extra argument would be a closure that lets you explicitly save the environment. So you can save the environment just after the provider.init and henceforth get the necessary environment for the destroy.

Do you want me to implement this?

[msimonin]

Here is what I had in mind. The destroy should also remove any container set up during the up phase (registry, cadvisor)

enos deploy
enos destroy

We could follow how kolla is implementing this.
For every role called by up.yml pass a variable action=up|destroy.
Each main.yml then include {{ action }}.yml and call the relevant play.

I agree that calling after a failed up could be problematic since the env is created at the end of the phase.

[rcherrueau]

Oh, you mean your title was about "destroying registry, cadvisor, ..." and not about "destroying provider resources"?

[msimonin]

'resources' is confusing :)

[rcherrueau]

😄