From c7fb1596329b65e4afdf3051d4e7266b67c66b79 Mon Sep 17 00:00:00 2001
From: Donavan Fritz <dfritz@netflix.com>
Date: Fri, 24 Apr 2026 23:11:47 -0500
Subject: [PATCH] agent: maintain NetworkUnavailable=False on owned nodes

When Calico shuts down on a flock-labeled node, calico-node sets
NetworkUnavailable=True with reason CalicoIsDown. Nothing replaces it,
so kubelet's NodeController applies node.kubernetes.io/network-
unavailable:NoSchedule and new pods can't land.

flock-agent now patches Status.Conditions every 60s with
NetworkUnavailable=False (reason=FlockReady). RBAC: nodes/status patch.

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
---
 deploy/install.yaml             |  3 ++
 deploy/rbac/serviceaccount.yaml |  3 ++
 pkg/agent/nodecondition.go      | 66 +++++++++++++++++++++++++++++++++
 pkg/agent/runtime_linux.go      |  6 +++
 4 files changed, 78 insertions(+)
 create mode 100644 pkg/agent/nodecondition.go

diff --git a/deploy/install.yaml b/deploy/install.yaml
index 5baf640..7fc587a 100644
--- a/deploy/install.yaml
+++ b/deploy/install.yaml
@@ -91,6 +91,9 @@ rules:
   - apiGroups: ["networking.k8s.io"]
     resources: ["networkpolicies"]
     verbs: ["get", "list", "watch"]
+  - apiGroups: [""]
+    resources: ["nodes/status"]
+    verbs: ["patch"]
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRoleBinding
diff --git a/deploy/rbac/serviceaccount.yaml b/deploy/rbac/serviceaccount.yaml
index d317e63..594ac4d 100644
--- a/deploy/rbac/serviceaccount.yaml
+++ b/deploy/rbac/serviceaccount.yaml
@@ -18,6 +18,9 @@ rules:
   - apiGroups: ["networking.k8s.io"]
     resources: ["networkpolicies"]
     verbs: ["get", "list", "watch"]
+  - apiGroups: [""]
+    resources: ["nodes/status"]
+    verbs: ["patch"]
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRoleBinding
diff --git a/pkg/agent/nodecondition.go b/pkg/agent/nodecondition.go
new file mode 100644
index 0000000..af06471
--- /dev/null
+++ b/pkg/agent/nodecondition.go
@@ -0,0 +1,66 @@
+package agent
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"time"
+
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/rest"
+)
+
+// keepNetworkAvailable maintains a NetworkUnavailable=False condition on
+// the node's status. Calico-node sets this False while it owns CNI; on
+// shutdown it sets it to True with reason CalicoIsDown, which adds the
+// node.kubernetes.io/network-unavailable taint and blocks new scheduling.
+// Once flock-agent is in charge, we own the condition.
+//
+// Re-applies every minute — heartbeat-style — so a stale condition from a
+// previous CNI is overwritten without an explicit transition.
+func keepNetworkAvailable(ctx context.Context, cfg *rest.Config, node string, logger *slog.Logger) {
+	cs, err := kubernetes.NewForConfig(cfg)
+	if err != nil {
+		logger.Warn("network-condition: kubernetes client", "err", err)
+		return
+	}
+	apply := func() {
+		now := metav1.Now()
+		patch := map[string]interface{}{
+			"status": map[string]interface{}{
+				"conditions": []corev1.NodeCondition{{
+					Type:               corev1.NodeNetworkUnavailable,
+					Status:             corev1.ConditionFalse,
+					Reason:             "FlockReady",
+					Message:            "flock-agent owns CNI on this node",
+					LastHeartbeatTime:  now,
+					LastTransitionTime: now,
+				}},
+			},
+		}
+		body, _ := json.Marshal(patch)
+		_, err := cs.CoreV1().Nodes().Patch(ctx, node, types.MergePatchType, body, metav1.PatchOptions{}, "status")
+		if err != nil {
+			logger.Warn("network-condition: patch failed", "err", err)
+			return
+		}
+	}
+	apply()
+	t := time.NewTicker(60 * time.Second)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			apply()
+		}
+	}
+}
+
+// silence unused-import warnings on non-Linux builds where this is unused.
+var _ = fmt.Sprintf
diff --git a/pkg/agent/runtime_linux.go b/pkg/agent/runtime_linux.go
index c507e38..5954660 100644
--- a/pkg/agent/runtime_linux.go
+++ b/pkg/agent/runtime_linux.go
@@ -53,6 +53,12 @@ func (s *Server) configureRuntime(ctx context.Context) error {
 		return fmt.Errorf("pod informer: %w", err)
 	}
 
+	// Keep NetworkUnavailable=False so the node.kubernetes.io/network-
+	// unavailable taint never gets re-applied. Calico's calico-node sets
+	// it on shutdown; without an owner replacing it, kubelet's controller
+	// taints the node and blocks scheduling.
+	go keepNetworkAvailable(ctx, s.restCfg, s.Node, s.Logger)
+
 	bird := &BirdManager{
 		NodeName:    s.Node,
 		ConfigPath:  "/etc/flock/bird/bird.conf",