Conversation
|
I have changed your target branch to 0.4.X. I am just working on a new branch workflow and I am documenting it. Sorry I don't know why that has just bombed your git history |
37f4db8 to
e1159cd
Compare
e1159cd to
bfd9306
Compare
|
Is the |
Yes and no We rely on the implicit fact Python comes in the image and I ran into this after I accidentally removed Python on a base image I was testing to avoid constantly re-downloading from upstream. Hence putting this role in (and fixing that image to match again) as it is the only hard-dependency Ansible has on VMs If we drop it then it won't break anything, but it's more a correctness thing that might guard against changes in the images in the future too |
bfd9306 to
7679441
Compare
This fixes a couple of issues: - An incorrect package name which isn't required for ncm - Quattor repos need to go into the correct location, else quattor-configure spma will cause duplicates - Ensure quattor-configure runs on first boot - Tidy image shouldn't run quattor. Especially since depending on archetype we'll just undo all the changes e.g. switching from cloud -> tier-1 on first boot. Ship an image that can bootstrap itself
Otherwise we clean our image, then pollute it again with logs and dnf package cache containing quattor steps as that runs after our configure_os_images step
If ssh reconnects but the VM's SSHd restarts (e.g. distro sync) this will hang the connection for a very long time based on the ssh caller's default which is (by-default) do not timeout. Instead if Ansible cannot do anything after 5 seconds over SSH reconnect and retry using the modules default 600 second limit for retries. This resolves the hang seen.
7679441 to
7a5615a
Compare
|
Re-tested with RL8 and RL9 images, correctly configures the packages using cloud-init and works on first boot |
2 participants
This fixes a couple of issues: