2018-04-28 20:26:12 +08:00
|
|
|
|
# Envoy mesh 教程
|
|
|
|
|
|
2018-12-11 20:40:34 +08:00
|
|
|
|
**注意:本书中的 Service Mesh 章节已不再维护,请转到 [istio-handbook](https://jimmysong.io/istio-handbook) 中浏览。**
|
|
|
|
|
|
2018-04-28 20:26:12 +08:00
|
|
|
|
本文是在 Kubernetes 集群中,使用 Envoy 来做 mesh,来为一个简单的使用 Python 编写的 Flask 应用程序做反向代理和负载均衡。
|
|
|
|
|
|
|
|
|
|
**注**:本教程中的示例来自 [envoy-steps](https://github.com/datawire/envoy-steps),本文中使用的所有的代码和 YAML 配置见 [envoy-tutorial](https://github.com/rootsongjc/envoy-tutorial)。
|
|
|
|
|
|
|
|
|
|
![Envoy Mesh架构图](../images/envoy-mesh-in-kubernetes.png)
|
|
|
|
|
|
|
|
|
|
## 前提条件
|
|
|
|
|
|
|
|
|
|
使用 [kubernetes-vagrant-centos-cluster](https://github.com/rootsongjc/kubernetes-vagrant-centos-cluster) 部署 kubernetes 集群,只要启动集群并安装了 CoreDNS 即可,无须安装其他插件。
|
|
|
|
|
|
|
|
|
|
## 部署应用
|
|
|
|
|
|
|
|
|
|
我们首先将应用部署到 Kubernetes 中。
|
|
|
|
|
|
|
|
|
|
部署 postgres 数据库。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl apply -f postgres
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
创建 usersvc 镜像。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
docker build -t jimmysong/usersvc:step1 .
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
部署 usersvc。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl apply -f usersvc
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
查看 uservc 的 ClusterIP 地址。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ kubectl get svc usersvc
|
|
|
|
|
kubectl get svc usersvc
|
|
|
|
|
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
|
|
|
|
|
usersvc ClusterIP 10.254.176.248 <none> 5000/TCP 11m
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
进到 node1 中访问该服务,因为我们要访问的是 ClusterIP,在我们自己的电脑上是无法直接访问的,所以进到虚拟机中操作。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ vagrant ssh node1
|
|
|
|
|
$ curl 10.254.176.248:5000
|
|
|
|
|
{
|
|
|
|
|
"hostname": "usersvc-7cf5bb9d85-9gx7w",
|
|
|
|
|
"msg": "user health check OK",
|
|
|
|
|
"ok": true,
|
|
|
|
|
"resolvedname": "172.33.10.7"
|
|
|
|
|
}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
尝试添加一个名为 `Alice` 的用户。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ curl -X PUT -H "Content-Type: application/json" \
|
|
|
|
|
-d '{ "fullname": "Alice", "password": "alicerules" }' \
|
|
|
|
|
10.254.176.248/user/alice
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
将会看到类似如下的输出。
|
|
|
|
|
|
|
|
|
|
```json
|
|
|
|
|
{
|
|
|
|
|
"fullname": "Alice",
|
|
|
|
|
"hostname": "usersvc-7cf5bb9d85-9gx7w",
|
|
|
|
|
"ok": true,
|
|
|
|
|
"resolvedname": "172.33.10.7",
|
|
|
|
|
"uuid": "EF43B475F65848C6BE708F436305864B"
|
|
|
|
|
}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
尝试再添加一个名为 `Bob` 的用户。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ curl -X PUT -H "Content-Type: application/json" \
|
|
|
|
|
-d '{ "fullname": "Bob", "password": "bobrules" }' \
|
|
|
|
|
10.254.176.248/user/bob
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
将会看到类似如下的输出。
|
|
|
|
|
|
|
|
|
|
```json
|
|
|
|
|
{
|
|
|
|
|
"fullname": "Bob",
|
|
|
|
|
"hostname": "usersvc-7cf5bb9d85-9gx7w",
|
|
|
|
|
"ok": true,
|
|
|
|
|
"resolvedname": "172.33.10.7",
|
|
|
|
|
"uuid": "6AC944E7D4254D9A811A82C0FDAC3046"
|
|
|
|
|
}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
当应用部署完毕后,我们该部署 edge envoy 了。
|
|
|
|
|
|
|
|
|
|
## 部署 edge envoy
|
|
|
|
|
|
|
|
|
|
部署 edge envoy 的方式很简单,执行下面的命令。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl apply -f edge-envoy
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
现在访问 edge envoy 是就可以路由到 `usersvc` 上的,当然直接访问 `usersvc` 也是可以的。
|
|
|
|
|
|
|
|
|
|
我们看下 edge-envoy 的 envoy 配置文件定义。
|
|
|
|
|
|
|
|
|
|
```json
|
|
|
|
|
{
|
|
|
|
|
"listeners": [
|
|
|
|
|
{
|
|
|
|
|
"address": "tcp://0.0.0.0:80",
|
|
|
|
|
"filters": [
|
|
|
|
|
{
|
|
|
|
|
"type": "read",
|
|
|
|
|
"name": "http_connection_manager",
|
|
|
|
|
"config": {
|
|
|
|
|
"codec_type": "auto",
|
|
|
|
|
"stat_prefix": "ingress_http",
|
|
|
|
|
"route_config": {
|
|
|
|
|
"virtual_hosts": [
|
|
|
|
|
{
|
|
|
|
|
"name": "backend",
|
|
|
|
|
"domains": ["*"],
|
|
|
|
|
"routes": [
|
|
|
|
|
{
|
|
|
|
|
"timeout_ms": 0,
|
|
|
|
|
"prefix": "/user",
|
|
|
|
|
"cluster": "usersvc"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
"filters": [
|
|
|
|
|
{
|
|
|
|
|
"type": "decoder",
|
|
|
|
|
"name": "router",
|
|
|
|
|
"config": {}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
],
|
|
|
|
|
"admin": {
|
|
|
|
|
"access_log_path": "/dev/null",
|
|
|
|
|
"address": "tcp://127.0.0.1:8001"
|
|
|
|
|
},
|
|
|
|
|
"cluster_manager": {
|
|
|
|
|
"clusters": [
|
|
|
|
|
{
|
|
|
|
|
"name": "usersvc",
|
|
|
|
|
"connect_timeout_ms": 250,
|
|
|
|
|
"type": "strict_dns",
|
|
|
|
|
"service_name": "usersvc",
|
|
|
|
|
"lb_type": "round_robin",
|
|
|
|
|
"features": "http2",
|
|
|
|
|
"hosts": [
|
|
|
|
|
{
|
|
|
|
|
"url": "tcp://usersvc:80"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
客户端访问 `edge-envoy` 的 `ClusterIP:8000/user/health` 就可以检查节点的健康状况。
|
|
|
|
|
|
|
|
|
|
## 部署 usersvc2
|
|
|
|
|
|
|
|
|
|
删除原来的 `usersvc`,部署第二版 `usersvc2`,它与原来的 `usersvc` 唯一不同的地方是在 `entrypoint` 中集成了 envoy,查看 `Dockerfile` 中指定的 `entrypoint.sh` 的内容便可知。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
#!/bin/sh
|
|
|
|
|
|
|
|
|
|
python /application/service.py &
|
|
|
|
|
/usr/local/bin/envoy -c /application/envoy.json
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
首先删除老的 `usersvc`。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl delete -f usersvc
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
使用下面的命令部署 `usersvc2`,它仍然使用 `usersvc` 这个 service 名称。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl apply -f usersvc2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
Envoy 以 out-of-process 的方式运行,对应用进程没有侵入性,也可以使用 sidecar 的方式运行,让 envoy 与 应用容器运行在同一个 pod 中。
|
|
|
|
|
|
|
|
|
|
增加 `usersvc2` 的实例个数。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl scale --replicas=3 deployment/usersvc
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
此时我们有 3 个 usersvc 实例,现在通过 `edge-envoy` 的 `ClusterIP:8000/user/health` 检查节点的健康状况时,是不是会轮询的访问到后端的的 `usersvc2` 的实例呢?
|
|
|
|
|
|
|
|
|
|
我们当初在 `edge-node` 的 `envoy.json` 中配置过 cluster 的,其中指定了 `lb_type` 为 `round_robin` 。
|
|
|
|
|
|
|
|
|
|
```json
|
|
|
|
|
"cluster_manager": {
|
|
|
|
|
"clusters": [
|
|
|
|
|
{
|
|
|
|
|
"name": "usersvc",
|
|
|
|
|
"connect_timeout_ms": 250,
|
|
|
|
|
"type": "strict_dns",
|
|
|
|
|
"service_name": "usersvc",
|
|
|
|
|
"lb_type": "round_robin",
|
|
|
|
|
"features": "http2",
|
|
|
|
|
"hosts": [
|
|
|
|
|
{
|
|
|
|
|
"url": "tcp://usersvc:80"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
而且该 `serivce_name` 也可以被 DNS 正确解析。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
root@usersvc-55b6857d44-gcg5c:/application# nslookup usersvc
|
|
|
|
|
Server: 10.254.0.2
|
|
|
|
|
Address: 10.254.0.2#53
|
|
|
|
|
|
|
|
|
|
Name: usersvc.envoy-tutorial.svc.cluster.local
|
|
|
|
|
Address: 10.254.123.166
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
**答案是否定的。**
|
|
|
|
|
|
|
|
|
|
虽然通过 DNS 可以正确的解析出 serivce 的 ClusterIP,但是负载均衡不再通过 kube-proxy 实现,所以不论我们访问多少次 `edge-envoy` 永远只能访问到一个固定的后端 `usersvc`。
|
|
|
|
|
|
|
|
|
|
## 服务发现服务 - SDS
|
|
|
|
|
|
|
|
|
|
Kubernetes 中的 DNS 可以发现所有 serivce 的 ClusterIP,但是 DNS 中不包括所有 endpoint 地址,我们需要一个 SDS(服务发现服务)来发现服务的所有的 endpoint,我们将修改 `lb_type`,使用 `sds` 替代 `strict_dns`。
|
|
|
|
|
|
|
|
|
|
执行下面的命令部署 SDS。
|
|
|
|
|
|
2018-05-25 14:21:04 +08:00
|
|
|
|
```bash
|
2018-04-28 20:26:12 +08:00
|
|
|
|
kubectl apply -f usersvc-sds
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
因为在添加了 SDS 之后需要修改 `edge-envoy` 中的 `envoy.josn` 配置,在 `clusters` 字段中增加 `sds` 信息,我们将所有的配置都写好了,重新打包成了镜像,我们需要先删除之前部署的 `edge-envoy`。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl delete -f edge-envoy
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
部署新的 `edge-envoy2`。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
kubectl apply -f edge-envoy2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
连续访问 `usersvc` 12 次看看输出结果如何。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
URL=http://172.17.8.101:30800/user/alice
|
|
|
|
|
for i in `seq 1 12`;do curl -s $URL|grep "resolvedname"|tr -d " "|tr -d ","|tr -d '"';done
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
我们可以看到类似如下的输出:
|
|
|
|
|
|
|
|
|
|
```ini
|
|
|
|
|
resolvedname:172.33.71.2
|
|
|
|
|
resolvedname:172.33.88.2
|
|
|
|
|
resolvedname:172.33.10.2
|
|
|
|
|
resolvedname:172.33.71.2
|
|
|
|
|
resolvedname:172.33.88.2
|
|
|
|
|
resolvedname:172.33.10.2
|
|
|
|
|
resolvedname:172.33.71.2
|
|
|
|
|
resolvedname:172.33.88.2
|
|
|
|
|
resolvedname:172.33.10.2
|
|
|
|
|
resolvedname:172.33.71.2
|
|
|
|
|
resolvedname:172.33.88.2
|
|
|
|
|
resolvedname:172.33.10.2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
再查看下 `usersvc` 服务的所有 pod 的 IP 地址。
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
$ kubectl get pod -l service=usersvc -o wide
|
|
|
|
|
NAME READY STATUS RESTARTS AGE IP NODE
|
|
|
|
|
usersvc-55b6857d44-mkfpv 1/1 Running 0 9m 172.33.88.2 node1
|
|
|
|
|
usersvc-55b6857d44-q98jg 1/1 Running 0 9m 172.33.71.2 node2
|
|
|
|
|
usersvc-55b6857d44-s2znk 1/1 Running 0 9m 172.33.10.2 node3
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
我们看到 round-robin 负载均衡生效了。
|
|
|
|
|
|
|
|
|
|
## 参考
|
|
|
|
|
|
|
|
|
|
- [Part 2: Deploying Envoy with a Python Flask webapp and Kubernetes](https://www.datawire.io/envoyproxy/envoy-flask-kubernetes/)
|
|
|
|
|
- [envoy-steps](https://github.com/datawire/envoy-steps)
|
|
|
|
|
- [kubernetes-vagrant-centos-cluster](https://github.com/rootsongjc/kubernetes-vagrant-centos-cluster)
|
2019-04-24 14:30:18 +08:00
|
|
|
|
- [envoy-tutorial](https://github.com/rootsongjc/envoy-tutorial)
|