From 789927333e11d3268d751bbd84ca728d84110411 Mon Sep 17 00:00:00 2001
From: Weihang Lo <weihang.lo@suse.com>
Date: Wed, 24 Nov 2021 17:19:18 +0800
Subject: [PATCH] Skip unreachable node during collecting NodeBundles

We currently skip nodes with unavailable network and not ready kubelet.
Nodes tainted with `node.kubernetes.io/unschedulable:NoSchedule` are tolerated.

Signed-off-by: Weihang Lo <weihang.lo@suse.com>
---
 pkg/manager/agent.go   |  5 +++++
 pkg/manager/manager.go | 15 +++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/pkg/manager/agent.go b/pkg/manager/agent.go
index 2f7193db..1ed26fbc 100644
--- a/pkg/manager/agent.go
+++ b/pkg/manager/agent.go
@@ -72,6 +72,11 @@ func (a *AgentDaemonSet) Create(image string, managerURL string) error {
 							Key:   types.DrainKey,
 							Value: "scheduling",
 						},
+						{
+							Key:      corev1.TaintNodeUnschedulable,
+							Operator: corev1.TolerationOpExists,
+							Effect:   corev1.TaintEffectNoSchedule,
+						},
 					},
 					Containers: []corev1.Container{
 						{
diff --git a/pkg/manager/manager.go b/pkg/manager/manager.go
index 888a8879..424a7725 100644
--- a/pkg/manager/manager.go
+++ b/pkg/manager/manager.go
@@ -15,6 +15,7 @@ import (
 	"github.com/rancher/wrangler/pkg/signals"
 	"github.com/sirupsen/logrus"
 
+	v1 "k8s.io/api/core/v1"
 	"k8s.io/client-go/rest"
 
 	"github.com/rancher/support-bundle-kit/pkg/manager/client"
@@ -323,7 +324,21 @@ func (m *SupportBundleManager) refreshNodes() error {
 	}
 
 	m.expectedNodes = make(map[string]string)
+
+NODE_LOOP:
 	for _, node := range nodes.Items {
+		for _, cond := range node.Status.Conditions {
+			switch cond.Type {
+			case v1.NodeReady:
+				if cond.Status != v1.ConditionTrue {
+					continue NODE_LOOP
+				}
+			case v1.NodeNetworkUnavailable:
+				if cond.Status == v1.ConditionTrue {
+					continue NODE_LOOP
+				}
+			}
+		}
 		m.expectedNodes[node.Name] = ""
 	}