Skip to content

Release Retro 25.3.0 #694

@Techassi

Description

@Techassi

Part of #686

After the 25.3.0 release, we will hold a release-retro to discuss what went well, and find out where we can improve.

  • Schedule the retro
  • Create action items during the retro

The task list below can capture items as we think of them.

Details
### Items for discussion
- [x] Having demos install to `default` really sucks. If they were deployed to another namespace, they would be easy to cleanup. We should add a parameter of the selected namespace. (there's a PR)
- [x] Demos should avoid installing cluster scoped resources (eg: Role/RoleBinding over ClusterRole/ClusterRoleBinding)
- [x] Demos install stable by default (now), but operators still install nightly by default. Stable should be the default.
- [x] Rework the build.yml workflow in operators (use our actions, deduplicate checks)
- [x] Add support to customize PR title in operator-templating rollout workflow
- [x] Add item to operator-templating pre-release.md issue template to also update cargo-cyclonedx and cargo-auditable (See action failures in https://github.com/stackabletech/docker-images/commit/980bdac2c51b591530038da930a40bac83a72100)
- [x] Pin rustup. Also, manually install the toolchain for sub projects (in Dockerfiles) because automatic installation is no longer supported, see: https://github.com/rust-lang/rustup/pull/3985 and https://github.com/stackabletech/docker-images/pull/1013
- [x] `make regenerate-nix` should be part of pre-commit in operators (so people don't have to remember to run it)
- [ ] Use issue types in templates (instead of epic, task, ... labels) - Lars stopped using issue types for _reasons_. Labels are still needed for now.
- [ ] Rename hadoop to hdfs in docker-images to avoid two names for the same thing. Also makes it easier to search for HDFS usages (e.g. when bumping the version in different operator repos). It could cause additional work. low priority
- [x] Rename docker-images repo to container-images (or product-images). low prio, but no objection.
- [x] Use the fast mirror for downloading sources of products instead of the slow mirror (add a flag for easy use). See https://github.com/stackabletech/docker-images/pull/1021
- [ ] Transform the download + upload new product version script to a GitHub workflow one can dispatch. @dervoeti has something cooking
- [x] Don't reuse tracking issues for product bumps. A lot of bumps were missed
- [x] Use the version definition list in docker-images as the source of truth instead of the spreadsheet. It provides versioning via Git, it can be used to build both the images and a support matrix for our docs. - Lars said yes, but needs to go through planning/refinement.
- [x] Make pre-commit in operators smarter about which changes each hook should run on
- [x] Remove old/stale branches across repos (branches with 0 commits ahead of main can go anytime)
- [x] Try to use `stackablectl demo in <demo> --release dev` instead of manually upgrading operators and patching product versions in stable -> nightly demo tests (`--force`?)
- [x] trino-taxi-data demo jobs should have an init container to wait for services to be ready before trying (link to e2e-security PR)
- [x] Some parts of the release scripts ignore failure (ie: `checks` with `set +e`)
- [x] Rustify (release related) stackable-utils scripts (on-site?)
- [x] E2E Testable release workflow
- [x] Split the Release Day (week) tasks in the template into _Change freeze_, _Release Candidate testing_, _Release tagging_
- [x] Do getting_started and demo testing after release branches are cut. See also https://github.com/stackabletech/issues/issues/703
- [x] Rename documentation branches to use `-` instead of `/` to be consistent with the other repos
- [x] Documentation `make-release-branch.sh` script needs to replace `HEAD` and `main` with the release branch (see: https://github.com/stackabletech/documentation/pull/725)
- [x] When demo release branch is created, `version` needs updating (Antora error about duplicate page detected)
- [x] Forgot to release stackablectl, seems to be missing from the tracking issue template (done in https://github.com/stackabletech/issues/pull/710)
- [x] stackablectl versionioning (uses SDP release versions, but is decoupled) - make a decision issue
- [x] Remove ci/release targets from operator Makefiles as they are done in actions/workflows (like in docker-images to build the images). Can keep the developer targets

Legend

  • ⏲ Time saver
  • 💥 Important
  • 🔽 Low priority

Action items

Deferred items

  • Use issue types in templates (instead of epic, task, ... labels)
  • Rename hadoop to hdfs in docker-images to avoid two names for the same thing. Also makes it easier to search for HDFS usages (e.g. when bumping the version in different operator repos). It could cause additional work. low priority
  • Transform the download + upload new product version script to a GitHub workflow one can dispatch. @dervoeti has something cooking

Metadata

Metadata

Assignees

No one assigned

    Labels

    release-retro/25.3.0Indicates this issue or PR is part of improvements resulting from the Release Retro 25.3.0

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions