Pulp Installers Meeting Minutes

Minutes
#22

Jun 1 Agenda

#23

Jun 8 Agenda

  • In the LDAP PR, I managed to make the extras variable usable even when users provide a string rather than a list.
  • In the LDAP PR, I did not do a case-insensitive search, even though extras are case-insensitive.
#24

Jun 15 Agenda

  • Research into other ansible installers:
    • ansible-freeipa
      • Has 1 role (e.g. “ipaclient”) per machine-type, but also has 1 hostgroup (e.g, “ipaclients”) per machine type.
      • Uses lots of custom python modules, and plugins
      • Custom python modules put the server application in the correct state after packages lay down files
      • installs RPM/deb packages
      • Has an uninstaller (`state|default(‘present’)``)
        • But wraps around a binary uninstall command
      • Specifies role short name rather than FQCN
    • foreman_basic_installer
      • Has 1 role (e.g. “foreman_smart_proxy”) per machine-type, but also has 1 hostgroup (e.g, “foreman_smart_proxy”) per machine type.
      • has a group_vars/all.yml.sample, with a long list of variables, that users can copy & modify
      • Has an inventory.sample file, which shows the hostgroups
      • Calls foreman-installer commands
      • Not idempotent, always restarts the webserver service for example.
    • AAP
      • Each webserver/pulp-host sets the content_origin to its own individual ansible_fqdn.
        • We could default it to the 1st host in pulp_webservers
    • brightcomputing/installer
      • Uses a table to describe variables.
      • specifies role’s FQCN
    • Team agreements:
      • Adopt hostgroups for the usability epic
  • lint failing on 3.18 branch
    • It’s as if GHA is like “I’m going to ignore existing lint errors, but not ones that are fixed on the other branch.”
    • Agreed: Just fix the lint errors
#25

Jun 22 Agenda

  • Need to answer these questions: https://issues.redhat.com/browse/AAP-4342
    • Only need help answering the question about what bug it will cause
    • Ask Ina
  • Status of fixing #1196 re-labeling mid-install
    • Can’t use ansible_facts.mounts because the options are excessively verbose, need to parse /etc/fstab instead
  • Status of fix for dev environments
#26

Jun 29 Agenda

  • [mikedep333] Focusing on the dev env for AH
    • Not blocking anyone’s work on the usability epic, am I?
  • vagrant for the old pulp2/pulp3 combo is broken
    • probably won’t need it soon because migration plugin will be EOL
    • look into whether hostmanager plugin breaks regular VMs
  • planning for the usability epic
    • desire for the installer to provide a “distribution” of pulp + plugins at specific versions
      • start a thread on discourse to discuss the installer providing default z-stream releases for plugins
  • updated centos 8stream vagrant images still not available
#27

Jul 13 Agenda

#28

Jul 20 Agenda

  • Making progress on special variable cleanup
    • Installer mimics the default behavior of pulpcore settings.py
#29

Jul 27 Agenda

  • I finished the special variable cleanup
  • In the middle of docs updates
    • Worry that the table will be too wide
  • In the middle of fixing 3.9 CI
    • For last 3.9 release, CI did not run. It seems to not run at all occassionally.
  • How to fix the ulimit bug on the operator
    • ulimit cannot be changed in the container
    • cgroups does not prescribe file limits (or other ulimits)
#30

Aug 3 Agenda

#31

Aug 17 Agenda

#32

Aug 24 Agenda

  • Figured out single container’s EL7 rootless compatibility issue
  • Figured out how to debug the initialization scripts for the single container
  • Doing lots of reading S6
  • Still having trouble with upgrade to v3, not sure whether to continue doing it or not
  • S6 is poorly documented and complex
    • “not simple at all”
  • Agreed: Use systemd instead of s6 if it doesn’t require any container permissions, and if it doesn’t slow down container startup
#33

Aug 31 Agenda

  • Ubuntu 22.04 support landed. Hurray!
  • New config format for S6
#34

Sep 7 Agenda

  • CentOS 9 and Postgres 13 upgrade in single container
    • includes upgrading existing database
      • No CI test for this though
    • Make discourse post
  • single container issues are the “production ready” task list
    • Some are marked prio-list. These are those being currently / soon worked on.
    • prio-list should be limited to 3 items at all times. Drop existing ones to add new ones.
  • Create new mkdocs site for single container
  • Container signing service landed
    • Includes cleaning up accidental logic from ansible signing service
    • file a bug, galaxy_ng’s example script is broken
#35

Sept 14 Agenda

#36

Sept 28 Agenda

  • Single container “production ready” coming along steadily
    • Made the decision to put the pulp process on a limited UID
  • ansible-lint keeps on getting stricter
  • RHEL support the installer (with repos to be enabled) was broken for months
#37

Oct 12 Agenda

  • [mikedep333] Been spending almost all my time fixing the CI, including for various repos
    • Includes some python test code
    • Need to check out the remainder of the repos
  • Could not reproduce the “double service” issue in oci-env
  • click env issue in single container is cluttering up CI logs
#38

Oct 19 Agenda

#39

Oct 26 Agenda

  • pulp-operator PR for the new images: https://github.com/pulp/pulp-operator/pull/698
    • fao89’s comment about not testing s6 images with the go operator
    • agreed: Only test the s6 images with the ansible operator. Since the Go operator will be a breaking change, and we do not want to create unnceccessary CI tests.
  • galaxy / galaxy-minimal image vs pulp-galaxy-ng image
    • ansible core CI (ansible/ansible on github) uses this image, but under the ansible namespace instead of pulp
    • Agreed: consolidate pulp-galaxy-ng to galaxy
#40

Nov 2 Agenda

  • status of compose development
  • How to announce breaking changes, or potentially breaking changes?
    • Put in development section of discourse?
    • Agreed: Ask in open floor
#41

Nov 16 Agenda