tripleorc is not available to debug a failed run

[Loganathan Parthipan]

My clean devtest failed at the overcloud formation. I wanted to source tripleorc to re-run overcloud alone wasn't available since it had never been written.

It appears that tripleorc gets written at the end of a successful devtest run. However, it would be useful if it gets written at the beginning.

[Ben Nemec (bnemec)]

tripleorc can't be written at the start because the env var configuration hasn't been done yet. It's possible we could write some intermediate versions for this sort of debugging, and I think that would address some of the arguments against tripleorc. Some investigation would be needed to see if those partial versions would be useful though. I suppose if nothing else they would take care of setting up TRIPLEO_ROOT and such.

[James Polley (tchaypo)]

I think the problem here is broader than just write-tripleorc not being available; devtest.sh as a whole seems to be designed with a one-shot all-or-nothing philosphy, with very little chance to pick up a failed run and continue with it.

I'm aware of two discussions happening at the moment to address this. https://etherpad.openstack.org/p/tripleo-devtest.sh-refactoring-blueprint is looking at updating devtest.sh to be more flexible; https://etherpad.openstack.org/p/tripleo-incubator-rationalise-ui is looking specifically at replacing the use of a wide array of environment variables with something that's more transparent.

[James Polley (tchaypo)]

https://review.openstack.org/#/c/108801/ is a partial attempt to address this.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Stephanie Miller (<email address hidden>) on branch: master
Review: https://review.openstack.org/108801
Reason: Changing directions - I don't think this is something we'll want to land now.