Testing Ansible and Friends together

tags: Design

Projects targetted

• ceph-ansible
• Ansible
• podman
• ansible-podman-collections
• Molecule [used in ansible role testing]

user story:

https://tree.taiga.io/project/tripleo-ci-board/us/1606

Goal

New releases of Ansible and friends does not break TripleO
Goal should me more descriptive here so that everyone will understand

Meeting Topics

• Current Status:
  • ceph-ansible with stable-4.0 branch and podman with latest master gatted with tripleo
  • No failures caught till now
• Do we want to gate Ansible and Friends? If yes
• Do we want to go ahead with the same way as we had done for third party testing?
  • Build the rpm and then feed into the CI
• What other jobs do we want to cover?
• Do we want to keep a sync what is going to come in next release of RHEL-8.x with the testing?
• Other questions?
• You mean here we want the third party job should run on recent stable branch/master branch of ceph-ansible?

Initial Goal

• Make sure ceph-ansible and ansible works together

Opinion from ceph-ansible team

• Summarry:
  • try to bump up ansible on the existing tripleo jobs and see if we got a successful deployment

Ceph-ansible test matrix

• check-rdo /check-github workflow
  ceph-ansible version <-> openstack-release <-> distro version
  master -> master -> centos-8
  stable-4.0 -> train -> centos-7

• testing with ansible-2.9 for openstack-regular pipeline
  ceph-ansible version <-> openstack-release <-> distro version -> ansible-version
  master -> master -> centos-8 -> ansible-2.9
  stable-4.0 -> train -> centos-7 -> 2.9

Things to fix based on current matrix without ansible

• tripleo-ceph-integration-master -> build rpm from master branch
  • currently running from stable-4.0 that needs to be fixed to master
  • Add sc1 and sc4 centos-8 based jobs which will be running from master branch
  • Make sure it gets triggered using check-rdo or check-github pr
  • Add these same jobs in regular pipeline
• Define a new job named tripleo-ceph-integration-train -> which will build rpm from stable-4.0 branch
  • It should be based on centos-7
  • Add sc1 and sc4 centos-7 based jobs which will be running from stable/train branch
  • Make sure it gets triggered using check-rdo or check-github pr
  • Add these same jobs in regular pipeline

Things to fix based on ansible-2.9

• add build rpm jobs
  • ceph-ansible master + ansible-2.9 -> centos-8
  • ceph-ansible stable-4.0 + ansible-2.9 -> centos-7
• Add scenario 1 and 4 job
  • master
    • ceph-ansible master + ansible-2.9 -> centos-8
  • train
    • ceph-ansible stable-4.0 + ansible-2.9 -> centos-7

Add below jobs:

Based on the discussion on IRC with John and Giulio, Here is what we
need to cover:
Jobs running when 'check-rdo' or 'check-github' comment is added to
ceph-ansible PR on

• ceph-ansible-4.0 branch
  • build rpm of ceph-ansible from stable-4.0 branch
    • for train
      • centos-7 rpm build job
      • rpm will get consumed in sc1 and sc4 on centos-7
• ceph-ansible stable-3.2 branch
  • build rpm of ceph-ansible stable-3.2 branch on centos-7
    • rpm will be get consumed in sc 1 and sc 4 job on centos-7
• • ceph-ansible-4.0 branch with ansible-2.9
  • build rpm of ceph-ansible from stable-4.0 branch + ansible -2.9
    • for train
      • centos-7 rpm build job
        • rpm will get consumed in sc1 and sc4 on centos-7
• ceph-ansible stable-3.2 branch + ansible-2.7
  • build rpm of ceph-ansible stable-3.2 branch on centos-7
    • rpm will be get consumed in sc 1 and sc 4 job on centos-7

Jobs running in regular or weekend pipeline

• Add all the above jobs here
• Add sc10 for master with ceph-ansible from stable-4.0 branch and ansible-2.9 on centos-8

Add jobs to test ceph-ansible and tripleo according to below test matrix

• acronym:
  ver = version
  sc = scenario10

• check-rdo /check-github workflow

  • Matrix-1:

    ceph-ansible ver <-> openstack-release <-> distro ver <-> ansible-ver <-> test
    stable-3.2 -> queens -> centos-7 -> -> sc001 and 004
    stable-3.2 -> queens -> centos-7 -> ansible-2.7 -> sc001 and 004

  • Matrix-2:

    ceph-ansible ver <-> openstack-release <-> distro ver <-> ansible-ver <-> test
    stable-4.0 -> train -> centos-7 -> -> sc001 and 004
    stable-4.0 -> train -> centos-7 -> ansible-2.9 -> sc001 and 004

  • Matrix-3:

    ceph-ansible ver <-> openstack-release <-> distro ver <-> ansible-ver <-> test
    stable-4.0-> master -> centos-8 -> -> sc001 and 004
    stable-4.0 -> master -> centos-8 -> ansible-2.9 -> sc001 and 004

  • Matrix-4:

    ceph-ansible ver <-> openstack-release <-> distro ver <-> ansible-ver <-> test
    stable-4.0-> master -> centos-8 -> ansible-2.9 -> sc10

• testing with openstack-regular pipeline

  • Add all the above added jobs into the openstack-regular pipline

• Note: Matrix-3 work is completed and Matrix 2 work is in progress.

Notes:

• Gate podman https://github.com/containers/libpod/tree/v1.4.2-stable

  • Need to fix the branch

• testing podman and ansible-podman-collection and ansible-podman-molecule driver jobs

  • change of podman should work with ansible-podman-collection

    • test the change with podman molecule driver

      • [sagi] can help on installing sf zuul ap on ansible-podman-collection
      • It is not used currently in tripleo deployment

    • Create userstory of podman , ansible-podman-collection and podman molecule driver.

• testing with Ansible and ceph-ansible

  • Create a job which will test the ansible-(stable-2.9 branch, Install it from pip) with ceph-ansible
    • In tripleo once ceph-ansible is getting installed
    • first remove ansible rpm
      • tell dnf not remove ansible deps
      • Install Ansible from pip
        • Hack on undercloud / overcloud to install Ansible from pip

• Need to find out escalation point of the failures from third party deps in future

  • need to find the ownership of failing jobs
    • Adding these jobs to grafana
    • Documenting Add job description if job fails, who owns the job and add the channel name to ping