MachinePool Maintenance

[friegger]

Summary

Today, when a bare-metal host backing a MachinePool needs maintenance (or fails unexpectedly),
there is no automated way to detect the failure, signal the intent, or gracefully evacuate Machines.
If the host shuts down, the VMs get killed, the API shows stale state, and operators must manually clean up.

This roadmap item introduces the building blocks for automated MachinePool lifecycle management.

---

Scope

In Scope

• Machine Pool Health
  • Pool Health Proposal Pool Health Proposal enhancements#6
  • Implement Pool Health Proposal Implement Pool Health Proposal ironcore#1472
• Cordon & Eviction
  • IEP Proposal: Machine Eviction Proposal enhancements#40
  • Implement Machine Eviction Proposal Implement Machine Eviction Proposal ironcore#1510
  • Documentation Added docs for machine eviction ironcore-dev.github.io#138
• MachinePool Lifecycle Controller
  • Implement MachinePool Lifecycle Controller Implement MachinePool Lifecycle Controller ironcore#1537
  • Documentation Added docs for MachinePool lifecycle controller ironcore-dev.github.io#139

---

Responsible Areas

• IaaS
• Compute

---

Contributors

• @friegger
• @lukasfrank

---

Acceptance Criteria

• a MachinePool Ready condition is marked as Unknown if it is not healthy
• a MachinePool is evicted once NoExecute is assigned
• a controller supports evicting a MachinePool as pre-drain activity for CAPI nodes

---

Action Items

• Assign labels (e.g., area/ironcore-api, kind/design)
• Set milestone (e.g., H1/2025)
• Assign dependent sub-issues in each required area
• Assign an owner to the issue using the GitHub "Assignee" field
• List all contributors in the "Contributors" section above
• Add this issue to the Roadmap project board