Closed wking closed 4 hours ago
@wking: This pull request references Jira Issue OCPBUGS-35994, which is valid. The bug has been moved to the POST state.
Requesting review from QA contact: /cc @jiajliu
The bug has been updated to refer to the pull request using the external bug tracker.
/cc
/cc
Service Delivery isn't very chatty about this, and re-reverts aren't hard if it comes to that.
/hold cancel
pre-merge verified /label qe-approved
@wking: This pull request references Jira Issue OCPBUGS-35994, which is valid.
Requesting review from QA contact: /cc @evakhoni
/lgtm
[APPROVALNOTIFIER] This PR is APPROVED
This pull-request has been approved by: petr-muller, wking
The full list of commands accepted by this bot can be found here.
The pull request process is described here
@wking: all tests passed!
Full PR test history. Your PR dashboard.
@wking: Jira Issue OCPBUGS-35994: All pull requests linked via external trackers have merged:
Jira Issue OCPBUGS-35994 has been moved to the MODIFIED state.
Current CI/QE capacity doesn't have space to support rollbacks, which require basically the same level of effort as supporting roll-forward updates, but which would never be used by as many clusters as the number of clusters that use roll-forward updates. In addition, paused MachineConfigPools and HyperShift's NodePools and similar allow for decoupling compute updates from control-plane updates. And control-plane updates have PodDisruptionBudgets and such to wedge gracefully if an issue comes up during updates that destabilizes the control plane. And if a control-plane update goes smoothly, but its behavior changes destabilize cluster workloads, ClusterVersion overrides and similar give an emergency safety valve to patch things up while admins wait for new releases that address the newly-discovered issues, while conditional updates allow the risk to be declared so additional clusters are warned before attempting an exposed update.
With the moderate operational benefit weighted against the substantial cost of CI/QE test coverage, this commit reverts the subcommand which I'd initially added in 73074c32ba (#1642) and gated in 9b2842a0c9 (#1764).