evrardjp · October 12, 2018 21:27 · evrardjp · Oct 12, 2018
diff --git a/osh-questions.txt b/osh-questions.txt
 Where to build and publish OSH images?
  New repo or not? Under which governance: LOCI or OSH?
  Push to DockerHub/Quay on a periodic basis?

 Where is the documentation about openstackhelm/kubeadm-aio:dev ?
  Is that system ubuntu only, or multi distro?
  
 Reviews: Could I get reviews on:
  https://review.openstack.org/#/q/project:%255Eopenstack/openstack-helm.*+owner:%22Jean-Philippe+Evrard+%253Cjean-philippe%2540evrard.me%253E%22++status:open+-DNM

 Discuss process to setup an environment that's close to zuul's environment? (Maybe I am wrong, please bear with me)
  I like the current ideas of the shell scripts, and the fact that start.sh is calling the gating plays
  I just am afraid we are setting a tangled mess that is not the developer experience.
  IMO the process should be clear like it is in the documentaion:
    1) Install requirements and k8s + helm (so it can be a BYO k8s) with a shell script
    2) Install OSH charts on it with the appropriate shell scritps
  Therefore the first step should not, IMO, be a shell script (005-deploy-k8s.sh) calling make dev-deploy, calling a script start.sh, installing ansible and running playbooks based on some conditionals.
  Why?
    Because you have either separated gates (running playbooks directly) and start.sh (which can be broken) OR you repeat yourself in gates (running plays then re-running said start script by running the initial .
    (Current state is repetition).
  My proposition:
    Gating still need a playbook. That playbook can be the script runner only, which will call a list of shell scripts, the first one (start.sh) should be common to all repos (and in -infra), and its focus should be on bootstrap-ansible + running the plays that can both work on gates and on developer environment (similar to what it is done today).  Opinion?
    all those playbooks
  Additional question: why is the makefile used for this bootstrap process? It seems more complex than it should.

 Makefiles and helm questions:
  What does this line actually do, when we know that "make all" before this step? https://github.com/openstack/openstack-helm-infra/blob/master/tools/deployment/developer/ceph/020-ingress.sh#L20
  Is there a way to be testing a helm chart without deploying? helm test/linting?  Should we have a make lint-all? Would you be okay to add this process in the zuul's linters jobs?
	Where to build and publish OSH images?
	New repo or not? Under which governance: LOCI or OSH?
	Push to DockerHub/Quay on a periodic basis?

	Where is the documentation about openstackhelm/kubeadm-aio:dev ?
	Is that system ubuntu only, or multi distro?

	Reviews: Could I get reviews on:
	https://review.openstack.org/#/q/project:%255Eopenstack/openstack-helm.*+owner:%22Jean-Philippe+Evrard+%253Cjean-philippe%2540evrard.me%253E%22++status:open+-DNM

	Discuss process to setup an environment that's close to zuul's environment? (Maybe I am wrong, please bear with me)
	I like the current ideas of the shell scripts, and the fact that start.sh is calling the gating plays
	I just am afraid we are setting a tangled mess that is not the developer experience.
	IMO the process should be clear like it is in the documentaion:
	1) Install requirements and k8s + helm (so it can be a BYO k8s) with a shell script
	2) Install OSH charts on it with the appropriate shell scritps
	Therefore the first step should not, IMO, be a shell script (005-deploy-k8s.sh) calling make dev-deploy, calling a script start.sh, installing ansible and running playbooks based on some conditionals.
	Why?
	Because you have either separated gates (running playbooks directly) and start.sh (which can be broken) OR you repeat yourself in gates (running plays then re-running said start script by running the initial .
	(Current state is repetition).
	My proposition:
	Gating still need a playbook. That playbook can be the script runner only, which will call a list of shell scripts, the first one (start.sh) should be common to all repos (and in -infra), and its focus should be on bootstrap-ansible + running the plays that can both work on gates and on developer environment (similar to what it is done today). Opinion?
	all those playbooks
	Additional question: why is the makefile used for this bootstrap process? It seems more complex than it should.

	Makefiles and helm questions:
	What does this line actually do, when we know that "make all" before this step? https://github.com/openstack/openstack-helm-infra/blob/master/tools/deployment/developer/ceph/020-ingress.sh#L20
	Is there a way to be testing a helm chart without deploying? helm test/linting? Should we have a make lint-all? Would you be okay to add this process in the zuul's linters jobs?