[epic] v1.0.0 Proposed release blockers

[joelanford]

Approved Release Blockers

Proposed Release Blockers

Non-blockers

• e2e error: already exists in namespace '' and cannot be managed by operator-controller #1307

Resolved

• Ensure rollouts are completed before starting a new upgrade. #1296
• versioned catalogd https interface RFC catalogd#426
• What is our public API? #1239
• Fix: e2e test for re-reconcile on catalog change is not valid #945
• The catalogd and operator-controller metrics services target each other #955
• Improve unit testing of reconcile() function. #879
• Bug: Bundle image source should use ext.Spec.InstallNamespace, not systemNamespace #944
• CRD Upgrade Safety preflight checks should check for incompatibility between served versions #1091
• Bug: When insufficient permissions exist to watch managed resources, reconciliation halts #1109
  • (Provided ServiceAccount) Follow-up: update dynamic cache to re-use cache.Cache instead of creating a new one every time #1025 (optimization) - associated with above issue
• 🐛 More API cleanups #1251
• catalogd never uses filesystem cache for image references with tags catalogd#392
• Possible bug: Do we properly handle mutations of spec.installNamespace? #961
• catalogmetadata client adds extreme latency to reconciliation #914
• Evaluate the use of conditions and reasoning in cluster extension controller #893
• Resolution fails when manually upgrading through legacy replaces chain beyond n+1 #1009
• ✨ Update to catalogd v0.17.0 to share a common certificate issuer #970
• Remove bundle.connection.config/insecureSkipTLSVerify annotation #921
• Bug: re-spun catalogd pods return 404 catalogd#288
• [epic] Improve kustomize layouts across OLMv1 components #967
• [Helm Integration] Evaluate if we need to store unpacked bundles in cache. #869
• operator-controller should not uninstall/rollback a failed install/upgrade #994
• [Bug] Reconciliation of cluster extension content not working on main branch #983
• Rukpak integration cleanup #987
• BUG: data too long issue rendering #923

[joelanford]

I took the liberty of moving a few of the release blockers to non-release blockers:

1. Fix: e2e test for re-reconcile on catalog change is not valid #945 is an e2e test. There is no indication today that the actual behavior related to reconciling ClusterExtensions due to ClusterCatalog changes is incorrect.
2. The catalogd and operator-controller metrics services target each other #955 does not change end-user facing APIs for functionality. It can be fixed after 1.0.0.
3. catalogmetadata client adds extreme latency to reconciliation #914 is a performance and UX issue, not an API or functionality change, so we can technically wait until after 1.0.0 to fix it. But I also still think that going out with this as a known issue would be a signal to users that we are unnecessarily rushing 1.0.0.

[tmshort]

Added #987, #879, #893. #869 as proposed release blockers

[tmshort]

And added #970 as a proposed release blocker.

[joelanford]

I don't think #869 is a release blocker because it is not a functional or API change. It is an optimization for reducing disk read/write/usage.

[everettraven]

@joelanford Just a heads up that I removed #1088 from this epic to be tracked in the v1 API review epic