metal-stack/machine-controller-manager-provider-metal

gardener machine controller manager for metal-stack

machine-controller-manager-provider-metal

Implements the OOT provider support for the Gardener machine-controller-manager.

machine-controller-manager-provider-metal's Issues

delete of all machines after mcm-restart

after mcm-restart all machines of a cluster were deleted:

I1109 14:56:54.687183       1 machine_safety.go:53] reconcileClusterMachineSafetyOrphanVMs: Start
I1109 14:56:56.009135       1 core.go:174] deleted machine "shoot--phjjbm--s3-test-01-group-0-77fc5-sr5zn" ("00000000-0000-0000-0000-ac1f6b7d7efa")
I1109 14:56:56.009192       1 machine_safety.go:297] SafetyController: Orphan VM found and terminated VM: shoot--phjjbm--s3-test-01-group-0-77fc5-sr5zn, metal:////00000000-0000-0000-0000-ac1f6b7d7efa
I1109 14:56:56.425097       1 core.go:174] deleted machine "shoot--phjjbm--s3-test-01-group-0-77fc5-qldng" ("00000000-0000-0000-0000-ac1f6b7d7e32")
I1109 14:56:56.425158       1 machine_safety.go:297] SafetyController: Orphan VM found and terminated VM: shoot--phjjbm--s3-test-01-group-0-77fc5-qldng, metal:////00000000-0000-0000-0000-ac1f6b7d7e32
I1109 14:56:56.842225       1 core.go:174] deleted machine "shoot--phjjbm--s3-test-01-group-0-77fc5-k5wp6" ("00000000-0000-0000-0000-ac1f6b7d7dc6")
I1109 14:56:56.842274       1 machine_safety.go:297] SafetyController: Orphan VM found and terminated VM: shoot--phjjbm--s3-test-01-group-0-77fc5-k5wp6, metal:////00000000-0000-0000-0000-ac1f6b7d7dc6
I1109 14:56:56.842441       1 machine_safety.go:68] reconcileClusterMachineSafetyOrphanVMs: End, reSync-Period: 30m0s
I1109 14:59:01.374674       1 machine_util.go:554] Conditions of Machine "shoot--phjjbm--s3-test-01-group-0-77fc5-qldng" with providerID "metal:///fra-equ01/00000000-0000-0000-0000-ac1f6b7d7e32" and backing node "shoot--phjjbm--s3-test-01-group-0-77fc5-qldng" are changing
W1109 14:59:01.374717       1 machine_util.go:562] Machine shoot--phjjbm--s3-test-01-group-0-77fc5-qldng is unhealthy - changing MachineState to Unknown

providerID is malformed

./machine-controller-manager-provider-metal/pkg/provider/core_util.go
39: return fmt.Sprintf("metal:///%s/%s", partition, machineID)

Logs emitted with V(2) do not appear in log output

Makes it harder to identify issues, e.g. #15.

Test migration path

We need to ensure that the migration to this sidecar and to the generic MachineClass resource works as expected such that workers do not get re-created. This would be catastrophic for our production environment.

Also we saw that in our test environment the MCM threw away seed worker nodes when the internet connection to the GKE control plane was defect.

• Test migration
  • Roll back environment to use old MCM deployment and all workers created with this old version
  • Update to new version and ensure no workers are thrown away
  • Machine classes were migrated successfully
• Cut the internet connection between seed and control plane for several hours
  • Ensure no workers are thrown away