nodehealth

type Controller ¶

type Controller struct {
	// contains filtered or unexported fields
}

Controller watches Node entities and marks sandboxes DEAD when their runner has been non-READY for longer than the grace period. This handles the case where a runner dies permanently (node failure, crash without recovery) and its sandboxes need to be cleaned up so the pool controller can create replacements.

The grace period exists because containers survive miren process restarts. A runner that's just restarting (upgrade, deploy) will come back and re-adopt its containers via reconcileSandboxesOnBoot(). We only want to intervene when the runner is truly gone.

Nodes with DISABLED status are intentionally excluded. DISABLED is set by Drain() during graceful shutdown, which handles its own sandbox cleanup. We don't want to race with that.

Implements controller.ReconcileControllerI[*compute_v1alpha.Node] Implements controller.DeletingReconcileController

func NewController(
	log *slog.Logger,
	eac *entityserver_v1alpha.EntityAccessClient,
) *Controller

func (c *Controller) Delete(ctx context.Context, id entity.Id) error

func (c *Controller) Init(ctx context.Context) error

func (c *Controller) Reconcile(ctx context.Context, node *compute_v1alpha.Node, meta *entity.Meta) error