layout: post title: k8s nginx ingress install date: 2021-01-05 18:54 +0800
2022-04-11 更新 k8s 1.23 版本寫法
WAN進來先用port forward走到虛擬IP的80,443
目前已經在cluster上架設keepalived + haproxy
haproxy listens on :80 :443 並開啟 proxy protocol
並自動將流量導向任意node的 :32080, :32443 使用 NodePort接收流量
cert-managerpanelpanel.foo.com 前後端分離defaultguest.bar.io一般使用者無法讀取secrets，但可以在自己的namespaec裡自由建立ingress
由於有多張wildcard，不能用default certificate選項指定
使用helm3進行安裝
helm repo add ingress-nginx https://kubernetes.github.io/ingress-nginx
helm repo update
helm install my-release ingress-nginx/ingress-nginx
ref: https://kubernetes.github.io/ingress-nginx/deploy/#using-helm
自訂values.yaml部分
controller:
config:
use-proxy-protocol: "true"
extraArgs:
http-port: 9080
https-port: 9443
ports:
http: 9080
https: 9443
targetPorts:
http: 9080
https: 9443
type: NodePort
nodePorts:
http: 32080
https:32443
helm install \
cert-manager jetstack/cert-manager \
--namespace ingress \
--version v1.7.2 \
--set installCRDs=true
我從gandi上買個幾個domain 使用上面提供的DNS server設定DNS記錄
接下來照著作
https://github.com/bwolf/cert-manager-webhook-gandi
首先確認 certificates 已經取得
kubectl get certificates
在namespace cert-manager使用這個ingress讓nginx解TLS流量
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
annotations:
nginx.ingress.kubernetes.io/ssl-redirect: "false"
name: foo-com-entry
namespace: cert-manager
spec:
ingressClassName: nginx
tls:
- hosts:
- "*.foo.com"
secretName: wildcard-foo-com-tls
rules:
- host: "*.foo.com"
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
annotations:
nginx.ingress.kubernetes.io/ssl-redirect: "false"
name: bar-io-entry
namespace: cert-manager
spec:
ingressClassName: nginx
tls:
- hosts:
- "*.bar.io"
secretName: wildcard-bar-io-tls
rules:
- host: "*.bar.io"
個別使用者可以用以下設定方式將特定host導到service上
這邊tls section都省略
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: guestbook
namespace: default
annotations: {}
spec:
ingressClassName: nginx
rules:
- host: guest.bar.io
http:
paths:
- path: /
pathType: Prefix
backend:
service:
name: frontend
port:
number: 80
nginx的rewrite功能我一直測試失敗，所以還是把prefix丟給service自己處理strip
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: panel
namespace: panel
annotations: {}
spec:
ingressClassName: nginx
rules:
- host: panel.foo.com
http:
paths:
- backend:
service:
name: panel-api
port:
number: 80
path: /api/
pathType: Prefix
- backend:
service:
name: panel-frontend
port:
number: 80
path: /
pathType: Prefix
走到default backend的時候由於沒有對應的cert
所以會回傳Kubernetes Fake Ingress Controller Certificates
nginx helm values.yaml 加上
controller:
config:
use-proxy-protocol: "true"
haproxy.cfg 修改
global
log /dev/log local0
log /dev/log local1 notice
daemon
#---------------------------------------------------------------------
# common defaults that all the 'listen' and 'backend' sections will
# use if not designated in their block
#---------------------------------------------------------------------
defaults
mode http
log global
option httplog
option dontlognull
option http-server-close
option forwardfor except 127.0.0.0/8
option redispatch
retries 1
timeout connect 5s
timeout client 7d
timeout server 7d
#---------------------------------------------------------------------
# apiserver frontend which proxys to the masters
#---------------------------------------------------------------------
frontend apiserver
bind *:8443
mode tcp
option tcplog
default_backend apiserver
#---------------------------------------------------------------------
# round robin balancing for apiserver
#---------------------------------------------------------------------
backend apiserver
mode tcp
option tcp-check
balance roundrobin
server k8s-node1 k8s-node1:6443 check
server k8s-node2 k8s-node2:6443 check
server k8s-node3 k8s-node3:6443 check
#---------------------------------------------------------------------
# ingress-http frontend which proxys to the k8s ingress controller
#---------------------------------------------------------------------
frontend ingress-http
bind *:80
mode tcp
option tcplog
default_backend ingress-http
#---------------------------------------------------------------------
# round robin balancing for ingress-http
#---------------------------------------------------------------------
backend ingress-http
mode tcp
option tcp-check
balance roundrobin
server k8s-node1 k8s-node1:32080 check send-proxy
server k8s-node2 k8s-node2:32080 check send-proxy
server k8s-node3 k8s-node3:32080 check send-proxy
#---------------------------------------------------------------------
# ingress-tls frontend which proxys to the k8s ingress controller
#---------------------------------------------------------------------
frontend ingress-tls
bind *:443
mode tcp
option tcplog
default_backend ingress-tls
#---------------------------------------------------------------------
# round robin balancing for ingress-tls
#---------------------------------------------------------------------
backend ingress-tls
mode tcp
option tcp-check
balance roundrobin
server k8s-node1 k8s-node1:32443 check send-proxy
server k8s-node2 k8s-node2:32443 check send-proxy
server k8s-node3 k8s-node3:32443 check send-proxy