fix 删除master/node流程

roles/kube-master/tasks/main.yml

@@ -91,6 +91,7 @@
     dest: /root/.kube/config
     regexp: "^    server"
     line: "    server: https://{{ inventory_hostname }}:6443"
+  tags: upgrade_k8s, restart_master
 
 - name: 以轮询的方式等待master服务启动完成
   command: "{{ bin_dir }}/kubectl get node"

roles/kube-node/defaults/main.yml

@@ -26,8 +26,15 @@ HARD_EVICTION: "memory.available<200Mi,nodefs.available<10%"
 BALANCE_ALG: "roundrobin"
 
 # 设置 APISERVER 地址
-KUBE_APISERVER: "{%- if groups['kube-master']|length > 1 -%} \
+KUBE_APISERVER: "{%- if inventory_hostname in groups['kube-master'] -%} \
+                     https://{{ inventory_hostname }}:6443 \
+                 {%- else -%} \
+                     {%- if groups['kube-master']|length > 1 -%} \
                          https://127.0.0.1:6443 \
                      {%- else -%} \
                          https://{{ groups['kube-master'][0] }}:6443 \
+                     {%- endif -%} \
                  {%- endif -%}"
+
+# 增加/删除 master 节点时，node 节点需要重新配置 haproxy 等
+MASTER_CHG: "no"

roles/kube-node/tasks/main.yml

@@ -25,7 +25,6 @@
     dest: /root/.kube/config
     regexp: "^    server"
     line: "    server: {{ KUBE_APISERVER }}"
-  when: "inventory_hostname not in groups['kube-master']"
 
 ##----------kubelet 配置部分--------------
 
@@ -46,15 +45,6 @@
         --embed-certs=true \
         --server={{ KUBE_APISERVER }} \
 	--kubeconfig=/etc/kubernetes/kubelet.kubeconfig"
-  when: "inventory_hostname not in groups['kube-master']"
-
-- name: 设置集群参数
-  shell: "{{ bin_dir }}/kubectl config set-cluster kubernetes \
-        --certificate-authority={{ ca_dir }}/ca.pem \
-        --embed-certs=true \
-        --server=https://{{ inventory_hostname }}:6443 \
-	--kubeconfig=/etc/kubernetes/kubelet.kubeconfig"
-  when: "inventory_hostname in groups['kube-master']"
 
 - name: 设置客户端认证参数
   shell: "{{ bin_dir }}/kubectl config set-credentials system:node:{{ inventory_hostname }} \
@@ -105,14 +95,6 @@
     dest: /etc/kubernetes/kube-proxy.kubeconfig
     regexp: "^    server"
     line: "    server: {{ KUBE_APISERVER }}"
-  when: "inventory_hostname not in groups['kube-master']"
-
-- name: 替换 kube-proxy.kubeconfig 的 apiserver 地址
-  lineinfile:
-    dest: /etc/kubernetes/kube-proxy.kubeconfig
-    regexp: "^    server"
-    line: "    server: https://{{ inventory_hostname }}:6443"
-  when: "inventory_hostname in groups['kube-master']"
 
 - name: 创建kube-proxy 服务文件
   template: src=kube-proxy.service.j2 dest=/etc/systemd/system/kube-proxy.service

roles/kube-node/tasks/node_lb.yml

@@ -35,3 +35,33 @@
   shell: systemctl start haproxy
   when: "groups['kube-master']|length > 1"
   tags: restart_lb
+
+# master 节点从1个增加到2个时候，需要修改如下配置
+# master 节点从2个减少到1个时候，也需要修改
+- block:
+  - name: 替换 kubeconfig 的 apiserver 地址
+    lineinfile:
+      dest: /root/.kube/config
+      regexp: "^    server"
+      line: "    server: {{ KUBE_APISERVER }}"
+  
+  - name: 替换 kubelet.kubeconfig 的 apiserver 地址
+    lineinfile:
+      dest: /etc/kubernetes/kubelet.kubeconfig
+      regexp: "^    server"
+      line: "    server: {{ KUBE_APISERVER }}"
+  
+  - name: 替换 kube-proxy.kubeconfig 的 apiserver 地址
+    lineinfile:
+      dest: /etc/kubernetes/kube-proxy.kubeconfig
+      regexp: "^    server"
+      line: "    server: {{ KUBE_APISERVER }}"
+
+  - name: restart kube-node service
+    service: name={{ item }} state=restarted
+    with_items:
+    - kubelet
+    - kube-proxy
+    ignore_errors: true
+  when: "MASTER_CHG == 'yes' and groups['kube-master']|length < 3"
+  tags: restart_lb

tools/12.delnode.yml

@@ -4,10 +4,6 @@
 - hosts: "{{ NODE_TO_DEL }}" 
   tasks:
   - name: fail info1
-    fail: msg="you CAN NOT delete the last member of etcd cluster!"
-    when: "groups['etcd']|length < 2 and NODE_TO_DEL in groups['etcd']"
-
-  - name: fail info2
     fail: msg="you CAN NOT delete the last member of kube-master!"
     when: "groups['kube-master']|length < 2 and NODE_TO_DEL in groups['kube-master']"
 
@@ -22,7 +18,7 @@
     DEL_LB: "yes"
   roles:
   - clean
-  task:
+  tasks:
   - name: 执行 kubectl delete node
     shell: "{{ bin_dir }}/kubectl delete node {{ NODE_TO_DEL }}"
     connection: local

tools/13.delmaster.yml

@@ -4,10 +4,6 @@
 - hosts: "{{ NODE_TO_DEL }}" 
   tasks:
   - name: fail info1
-    fail: msg="you CAN NOT delete the last member of etcd cluster!"
-    when: "groups['etcd']|length < 2 and NODE_TO_DEL in groups['etcd']"
-
-  - name: fail info2
     fail: msg="you CAN NOT delete the last member of kube-master!"
     when: "groups['kube-master']|length < 2 and NODE_TO_DEL in groups['kube-master']"
 
@@ -23,7 +19,7 @@
     DEL_LB: "yes"
   roles:
   - clean
-  task:
+  tasks:
   - name: 执行 kubectl delete node
     shell: "{{ bin_dir }}/kubectl delete node {{ NODE_TO_DEL }}"
     delegate_to: "{{ groups['kube-node'][0] }}"

tools/easzctl

@@ -107,7 +107,7 @@ function add-master() {
     ansible-playbook $BASEPATH/tools/03.addmaster.yml -e NODE_TO_ADD=$1 || { sed -i "/$1 NEW_MASTER=yes/d" $BASEPATH/hosts; return 2; }
 
     # reconfigure and restart the haproxy service on 'kube-node' nodes
-    ansible-playbook $BASEPATH/05.kube-node.yml -t restart_lb || { echo "[ERROR] Failed to restart the haproxy service on 'kube-node' nodes!"; return 2; }
+    ansible-playbook $BASEPATH/05.kube-node.yml -t restart_lb -e MASTER_CHG=yes || { echo "[ERROR] Failed to restart the haproxy service on 'kube-node' nodes!"; return 2; }
 
     # save current cluster context if needed
     [ -f "$BASEPATH/.cluster/current_cluster" ] && save_context
@@ -160,7 +160,7 @@ function del-node() {
     [[ $1 =~ ^(2(5[0-5]{1}|[0-4][0-9]{1})|[0-1]?[0-9]{1,2})(\.(2(5[0-5]{1}|[0-4][0-9]{1})|[0-1]?[0-9]{1,2})){3}$ ]] || { echo "[ERROR] Invalid ip address!"; return 2; }
 
     # 
-    ansible-playbook $BASEPATH/tools/12.delnode.yml -e ETCD_TO_DEL=$1 || { echo "[ERROR] Failed to delete 'kube-node': $1!"; return 2; }
+    ansible-playbook $BASEPATH/tools/12.delnode.yml -e NODE_TO_DEL=$1 || { echo "[ERROR] Failed to delete 'kube-node': $1!"; return 2; }
 
     # remove node in ansible hosts
     sed -i '/^\[kube-node/,/^\[harbor/{/^'"$1"'[^0-9]*$/d}' $BASEPATH/hosts
@@ -175,7 +175,7 @@ function del-master() {
     [[ $1 =~ ^(2(5[0-5]{1}|[0-4][0-9]{1})|[0-1]?[0-9]{1,2})(\.(2(5[0-5]{1}|[0-4][0-9]{1})|[0-1]?[0-9]{1,2})){3}$ ]] || { echo "[ERROR] Invalid ip address!"; return 2; }
 
     # 
-    ansible-playbook $BASEPATH/tools/13.delmaster.yml -e ETCD_TO_DEL=$1 || { echo "[ERROR] Failed to delete 'kube-master': $1!"; return 2; }
+    ansible-playbook $BASEPATH/tools/13.delmaster.yml -e NODE_TO_DEL=$1 || { echo "[ERROR] Failed to delete 'kube-master': $1!"; return 2; }
 
     # remove node in ansible hosts
     sed -i '/^\[kube-master/,/^\[kube-node/{/^'"$1"'[^0-9]*$/d}' $BASEPATH/hosts
@@ -183,6 +183,9 @@ function del-master() {
     # reconfig kubeconfig in ansible controller
     ansible-playbook $BASEPATH/01.prepare.yml -t create_kctl_cfg 
 
+    # reconfigure and restart the haproxy service on 'kube-node' nodes
+    ansible-playbook $BASEPATH/05.kube-node.yml -t restart_lb -e MASTER_CHG=yes || { echo "[ERROR] Failed to restart the haproxy service on 'kube-node' nodes!"; return 2; }
+
     # save current cluster context if needed
     [ -f "$BASEPATH/.cluster/current_cluster" ] && save_context
     return 0