osgeo-debian-11 LXD image on osgeo7 refers to a Debian 10 instead

[strk]

I've found today that the osgeo-debian-11 image on osgeo7 (didn't check other LXD hosts) is really pointing at a Debian 10 image. We should fix that. Dropping the alias/image could be enough for next deploy to recreate it.

[strk]

osgeo9 also has the bogus alias, so maybe that's the origin of the problem. osgeo4, osgeo3 and osgeo8 don't have that alias.

[robe]

I think it's debian11 but it originated from a debian10 which I upgraded to debian11.

I think for some reason when I copied over the image from osgeo9, it reverted back to the description of the original image. Didn't look into why that happened.

[strk]

Is there any reason why those images are not fresh debian11 ? The reason why I thought they were debian10 was because of the /etc/issue file stating it's a debian 10

[robe]

No particular reason aside from I was too lazy to reset up ldap auth. I think Alex said that would be a good role to have in ansible. Though the setup I think it's slightly different between OS.

[strk]

I agree the ldap auth should be done with an ansible role, ansible can handle OS differences

[strk]

Another sign of the debian-11 image NOT being Debian-11 is this:

  msg: No package matching 'python-is-python3' is available

That package is known to be available in Debian 11

The source.list file in a container created from osgeo-debian-11 still points at buster rather than bullseye

[strk]

I'd drop the image and recreate it from plain upstream images:debian-11 if you agree with this, we want the LDAP setup to be done with ansible anyway - is there a documentation somewhere about how you enabled the LDAP auth setup ?

[strk]

I've renamed the osgeo-debian-11 image to osgeo-debian-11-upgraded-by-10-see-trac2735 and will continue from there, creating a new osgeo-debian-11 image being an updated upstream debian-11 image

[strk]

It turns out it was ansible-deployment responsible for adding an osgeo-debian-11 alias while pointing at a debian-10 image. I fixed that with ​https://git.osgeo.org/gitea/sac/ansible-deployment/commit/dfeb4fc3125218913088b1c96ba4a58088573384

Before closing this out I shall fix all other LXD hosts too

[strk]

osgeo9 also fixed, by a new deploy from ansible, after renaming the bogux image to osgeo-debian-11-bogus-see-trac2735 (those images, both on osgeo9 and osgeo7, can be deleted now)

[robe]

I rebuilt the osgeo-debian-11 using a fresh debian-11 and installing the ldap osgeo stuff in. Script is here - ​https://git.osgeo.org/gitea/sac/ansible-deployment/src/branch/master/deployment/roles/images/tasks/build-osgeo-debian-11.yml

I'm rebuilding again to point the debian repo to debian.osuosl.org

I also still have to figure out how to put in a description on the images with ansible to properly label it as having OSGeo LDAP pre-installed.

As part of the lxd host role for production hosts, I do have it copy over the osgeo-debian-11 and osgeo-ubuntu-20 to the local host

[robe]

I was able to script addition of the description with below. A bit hacky but it worked. Basically it outputs the current image details with lxc image show and then seds that to inject with OSGeo LDAP SSH and then pipe's the new yml stuff with the new version using lxc image edit.

It's too bad there is no lxc add alias like thing for description.

- name: Add description to image
  shell: >
        lxc image show osgeo-debian-11 |
        sed 's/Debian bullseye amd64/Debian bullseye amd64 with OSGeo LDAP SSH /g' |
        lxc image edit osgeo-debian-11