Improve Conditions and Terminal errors

[EmilienM]

/kind feature

Context and Background

As part of the initiative to improve status reporting in Cluster API (CAPI) resources, significant changes will be introduced to how resource statuses are handled in the Cluster API Provider for OpenStack (CAPO).

One major change involves phasing out the FailureReason and FailureMessage fields in favor of leveraging Kubernetes Conditions to encapsulate terminal failures and lifecycle statuses. Terminal failures, though unique to CAPI, can be effectively communicated through well-defined conditions, using explicit type and reason values to represent fatal issues. This shift aligns CAPO with Kubernetes conventions and ensures that error states are consistently and clearly conveyed.

Key Updates and Behavior Changes

1. Handling Terminal Failures with Conditions

• Terminal failures will now be represented as conditions, providing clear, human-readable messages and actionable reasons.
• CAPO controllers will interpret these conditions to stop reconciling objects in a terminal state.
• If the failure is transient, users can manually clear the fatal condition to allow reconciliation to restart.

2. Lifecycle Management via Conditions

• Non-Recoverable Conditions: Objects with fatal conditions (e.g., unrecoverable infrastructure issues) will no longer be reconciled.
• Temporary Conditions: Objects with transient issues will continue reconciliation until either resolved or escalated to a fatal state.

3. Immutable vs. Mutable Resource Behavior

• Immutable Resources (OpenStackMachine, OpenStackServer): Once fully ready, these resources become immutable. Manual changes to the underlying OpenStack resources will not update their status in CAPO (e.g., replacing a server port).
• Mutable Resources (OpenStackCluster): These resources may experience condition changes, reflecting updates or failures after modification (e.g., issues arising from adding a security group while the Neutron API is unresponsive).

Known Issues and Areas for Improvement

Several existing issues highlight gaps in handling terminal failures or reflect inconsistent status behavior. This enhancement will address the following key issues:

• Issue #2146: Terminal failures are either not identified or incorrectly reported.
• Issue #2185: Missing conditions in critical resource workflows.
• Issue #2264: Inconsistent handling of fatal errors in OpenStackMachine.
• Issue #2265: Status fields are not aligned with the proposed lifecycle management.

Summary

By aligning CAPO with CAPI’s improved status reporting and transitioning to a condition-driven model, this enhancement will:

• Provide clearer, more actionable resource statuses.
• Reduce ambiguity in handling terminal failures.
• Improve lifecycle management for immutable and mutable resources.
• Address existing gaps and inconsistencies in error reporting.