Consul 是基于 GO 语言开发的开源工具,主要面向分布式,服务化的系统提供服务注册、服务发现和配置管理的功能。Consul 提供服务注册/发现、健康检查、Key/Value存储、多数据中心和分布式一致性保证等功能。Prometheus 通过 Consul 可以很方便的实现服务自动发现和维护,同时 Consul 支持分布式集群部署,将大大提高了稳定性,通过 Prometheus 跟 Consul 集群二者结合起来,能够高效的进行数据维护同时保证系统稳定。
本次演示环境,我是在虚拟机上安装 Linux 系统来执行操作,以下是安装的软件及版本:
注意:这里为了方便启动 Prometheus,我使用 Docker 方式启动,所以本机需要安装好 Docker 环境,这里忽略 Docker 的安装过程。其中 Prometheus 安装配置,可以参照之前文章 Prometheus 监控报警系统 AlertManager 之邮件告警,这里着重介绍一下如何配置 Consul 分布式集群,以及使用 nginx 来负载均衡 Consul 集群,最后分别验证配置 Prometheus 基于 Consul 集群来实现自动服务发现。
Consul 单机安装很方便,官网 提供各个系统版本二进制安装包,解压安装即可,可以参照之前文章 Prometheus 通过 consul 实现自动服务发现 文章来安装。这里我们要搭建 Consul 分布式集群,既然是分布式集群,那么肯定至少得部署到三台机器上,组成一个集群,苦于手上没有那么多的机器,我们只能在一台机器上部署三个 Consul 服务来模拟 “分布式” 集群,Consul 默认以 client 方式启动,这里我们采用 Server 方式来启动实例,通过不同的端口来区分不同的服务,集群实例如下:
下载最新版二进制安装包,解压后并拷贝到 /usr/local/bin
目录下
$ cd /root/prometheus/consul
$ wget https://releases.hashicorp.com/consul/1.6.1/consul_1.6.1_linux_amd64.zip
$ unzip consul_1.6.1_linux_amd64.zip
$ mv consul /usr/local/bin
集群启动三个实例,这里因为在一台机器上,直接使用 consul anget -server -bootstrap-expect=3 -data-dir=xxx...
命令来启动,稍显麻烦且不直观,这里我们采用配置文件的方式分别启动实例。配置 node1 实例,新建配置文件 consul01.json
如下 :
{
"datacenter": "dc1",
"data_dir": "/root/prometheus/consul/consul01",
"log_level": "INFO",
"server": true,
"node_name": "node1",
"ui": true,
"bind_addr": "172.30.12.100",
"client_addr": "0.0.0.0",
"advertise_addr": "172.30.12.100",
"bootstrap_expect": 3,
"ports":{
"http": 8500,
"dns": 8600,
"server": 8300,
"serf_lan": 8301,
"serf_wan": 8302
}
}
说明一下参数:
后台启动一下 node1 Consul 服务,执行如下命令启动:
nohup consul agent -config-dir=consul01.json > ./consul01.log 2>&1 &
查看服务启动日志会发现如下输出:
ootstrap_expect > 0: expecting 3 servers
==> Starting Consul agent...
Version: 'v1.6.1'
Node ID: '18450af8-64a3-aa0f-2a6c-4c2fc6e35a4b'
Node name: 'node1'
Datacenter: 'dc1' (Segment: '<all>')
Server: true (Bootstrap: false)
Client Addr: [0.0.0.0] (HTTP: 8500, HTTPS: -1, gRPC: -1, DNS: 8600)
Cluster Addr: 172.30.12.100 (LAN: 8301, WAN: 8302)
Encrypt: Gossip: false, TLS-Outgoing: false, TLS-Incoming: false, Auto-Encrypt-TLS: false
==> Log data will now stream in as it occurs:
2020/03/30 17:03:18 [INFO] raft: Initial configuration (index=0): []
2020/03/30 17:03:18 [INFO] raft: Node at 172.30.12.100:8300 [Follower] entering Follower state (Leader: "")
2020/03/30 17:03:18 [INFO] serf: EventMemberJoin: node1.dc1 172.30.12.100
2020/03/30 17:03:18 [INFO] serf: EventMemberJoin: node1 172.30.12.100
2020/03/30 17:03:18 [INFO] agent: Started DNS server 0.0.0.0:8600 (udp)
2020/03/30 17:03:18 [INFO] consul: Handled member-join event for server "node1.dc1" in area "wan"
2020/03/30 17:03:18 [INFO] consul: Adding LAN server node1 (Addr: tcp/172.30.12.100:8300) (DC: dc1)
2020/03/30 17:03:18 [INFO] agent: Started DNS server 0.0.0.0:8600 (tcp)
2020/03/30 17:03:18 [INFO] agent: Started HTTP server on [::]:8500 (tcp)
2020/03/30 17:03:18 [INFO] agent: started state syncer
==> Consul agent running!
==> Newer Consul version available: 1.7.2 (currently running: 1.6.1)
2020/03/30 17:03:23 [WARN] raft: no known peers, aborting election
2020/03/30 17:03:25 [ERR] agent: failed to sync remote state: No cluster leader
我们会看到类似 [ERR] agent: failed to sync remote state: No cluster leader
这样的报错,这是因为目前还没有任何其他实例加入,无法构成一个集群。接下来,按照上边的配置,启动第二个实例 node2,并加入到 node1 集群中,新建配置文件 consul02.json
如下 :
{
"datacenter": "dc1",
"data_dir": "/root/prometheus/consul/consul02",
"log_level": "INFO",
"server": true,
"node_name": "node2",
"ui": true,
"bind_addr": "172.30.12.100",
"client_addr": "0.0.0.0",
"advertise_addr": "172.30.12.100",
"bootstrap_expect": 3,
"ports":{
"http": 8510,
"dns": 8610,
"server": 8310,
"serf_lan": 8311,
"serf_wan": 8312
}
}
注意:由于这里是一台服务器上启动了多个实例,所以必须修改各个服务端口号,否则会报错,这里新启动的 Consul 服务端口为 8510
。启动 node2 Consul 服务,并加入到 node1 Consul 服务中组成集群,启动命令如下:
nohup consul agent -config-dir=consul02.json -join 172.30.12.100:8301 > ./consul02.log 2>&1 &
启动完毕后,此时会发现 node1 的日志输出中会打印如下日志:
2020/03/30 17:14:54 [INFO] serf: EventMemberJoin: node2 172.30.12.100
2020/03/30 17:14:54 [INFO] consul: Adding LAN server node2 (Addr: tcp/172.30.12.100:8310) (DC: dc1)
2020/03/30 17:14:54 [INFO] serf: EventMemberJoin: node2.dc1 172.30.12.100
2020/03/30 17:14:54 [INFO] consul: Handled member-join event for server "node2.dc1" in area "wan"
可以看到 node2 成功加入到 node1 集群,不过接下来它还会一直报错 agent: failed to sync remote state: No cluster leader
,这是为什么?日志明明已经显示 node2 成功加入了,怎么还一直报错呢?这是因为我们创建时指定了 bootstrap_expect
为 3
,那么就必须至少存在 3 个实例才能组成该 Consul 集群。那么,依照以上步骤,继续启动 node3 Consul 服务并添加到 node1 集群中,新建配置文件 consul03.json
如下:
{
"datacenter": "dc1",
"data_dir": "/root/prometheus/consul/consul03",
"log_level": "INFO",
"server": true,
"node_name": "node3",
"ui": true,
"bind_addr": "172.30.12.100",
"client_addr": "0.0.0.0",
"advertise_addr": "172.30.12.100",
"bootstrap_expect": 3,
"ports":{
"http": 8520,
"dns": 8620,
"server": 8320,
"serf_lan": 8321,
"serf_wan": 8322
}
}
启动 node3 Consul 实例命令如下:
nohup consul agent -config-dir=consul03.json -join 172.30.12.100:8301 > ./consul03.log 2>&1 &
此时,我们会发现 node1 日志输出中会打印如下日志:
2020/03/30 17:24:39 [INFO] serf: EventMemberJoin: node3 172.30.12.100
2020/03/30 17:24:39 [INFO] consul: Adding LAN server node3 (Addr: tcp/172.30.12.100:8320) (DC: dc1)
2020/03/30 17:24:39 [INFO] consul: Found expected number of peers, attempting bootstrap: 172.30.12.100:8300,172.30.12.100:8310,172.30.12.100:8320
2020/03/30 17:24:39 [INFO] serf: EventMemberJoin: node3.dc1 172.30.12.100
2020/03/30 17:24:39 [INFO] consul: Handled member-join event for server "node3.dc1" in area "wan"
2020/03/30 17:24:40 [WARN] raft: Heartbeat timeout from "" reached, starting election
2020/03/30 17:24:40 [INFO] raft: Node at 172.30.12.100:8300 [Candidate] entering Candidate state in term 2
2020/03/30 17:24:40 [INFO] raft: Election won. Tally: 2
2020/03/30 17:24:40 [INFO] raft: Node at 172.30.12.100:8300 [Leader] entering Leader state
2020/03/30 17:24:40 [INFO] raft: Added peer b7212be5-7d33-a5b6-0ca8-522f05ab9eeb, starting replication
2020/03/30 17:24:40 [INFO] raft: Added peer 0afe6fcb-3a07-e443-3802-f822ae9df422, starting replication
2020/03/30 17:24:40 [INFO] consul: cluster leadership acquired
2020/03/30 17:24:40 [INFO] consul: New leader elected: node1
2020/03/30 17:24:40 [WARN] raft: AppendEntries to {Voter 0afe6fcb-3a07-e443-3802-f822ae9df422 172.30.12.100:8320} rejected, sending older logs (next: 1)
2020/03/30 17:24:40 [INFO] raft: pipelining replication to peer {Voter 0afe6fcb-3a07-e443-3802-f822ae9df422 172.30.12.100:8320}
2020/03/30 17:24:40 [WARN] raft: AppendEntries to {Voter b7212be5-7d33-a5b6-0ca8-522f05ab9eeb 172.30.12.100:8310} rejected, sending older logs (next: 1)
2020/03/30 17:24:40 [INFO] raft: pipelining replication to peer {Voter b7212be5-7d33-a5b6-0ca8-522f05ab9eeb 172.30.12.100:8310}
2020/03/30 17:24:40 [INFO] consul: member 'node1' joined, marking health alive
2020/03/30 17:24:40 [INFO] consul: member 'node2' joined, marking health alive
2020/03/30 17:24:40 [INFO] consul: member 'node3' joined, marking health alive
2020/03/30 17:24:42 [INFO] agent: Synced node info
node3 成功加入到 node1 集群,整个集群已选出 node1 作为新的 Leader,此时集群状态为 health,那么整个 “分布式” 集群就搭建完毕了。此时,浏览器访问以下任意一个 http://127.0.0.1:8500
、http://127.0.0.1:8510
、http://127.0.0.1:8520
地址,均可打开 Consul Web 管理页面。
我们也可以通过命令行查看集群状态、以及集群成员状态。
# 查看集群状态
$ consul operator raft list-peers
Node ID Address State Voter RaftProtocol
node1 18450af8-64a3-aa0f-2a6c-4c2fc6e35a4b 172.30.12.100:8300 leader true 3
node2 b7212be5-7d33-a5b6-0ca8-522f05ab9eeb 172.30.12.100:8310 follower true 3
node3 0afe6fcb-3a07-e443-3802-f822ae9df422 172.30.12.100:8320 follower true 3
# 查看集群成员状态
$ consul members
Node Address Status Type Build Protocol DC Segment
node1 172.30.12.100:8301 alive server 1.6.1 2 dc1 <all>
node2 172.30.12.100:8311 alive server 1.6.1 2 dc1 <all>
node3 172.30.12.100:8321 alive server 1.6.1 2 dc1 <all>
接下来,我们配置 Prometheus 来使用 Consul 集群来实现自动服务发现,目的就是能够将添加的服务自动发现到 Prometheus 的 Targets 中,详细配置说明可以参考之前文章 Prometheus 通过 consul 实现自动服务发现 中的配置,在修改 Prometheus 配置之前,我们需要往 Consul 集群中注册一些数据。首先,我们注册一个 node-exporter-172.30.12.100
的服务,新建 consul-1.json
如下:
{
"ID": "node-exporter",
"Name": "node-exporter-172.30.12.100",
"Tags": [
"node-exporter"
],
"Address": "172.30.12.100",
"Port": 9100,
"Meta": {
"app": "spring-boot",
"team": "appgroup",
"project": "bigdata"
},
"EnableTagOverride": false,
"Check": {
"HTTP": "http://172.30.12.100:9100/metrics",
"Interval": "10s"
},
"Weights": {
"Passing": 10,
"Warning": 1
}
}
# 调用 API 注册服务
$ curl --request PUT --data @consul-1.json http://172.30.12.100:8500/v1/agent/service/register?replace-existing-checks=1
然后,我们再注册一个 cadvisor-exporter-172.30.12.100
的服务,新建 consul-2.json
并执行如下命令:
{
"ID": "cadvisor-exporter",
"Name": "cadvisor-exporter-172.30.12.100",
"Tags": [
"cadvisor-exporter"
],
"Address": "172.30.12.100",
"Port": 8080,
"Meta": {
"app": "docker",
"team": "cloudgroup",
"project": "docker-service"
},
"EnableTagOverride": false,
"Check": {
"HTTP": "http://172.30.12.100:8080/metrics",
"Interval": "10s"
},
"Weights": {
"Passing": 10,
"Warning": 1
}
}
# 调用 API 注册服务
$ curl --request PUT --data @consul-2.json http://172.30.12.100:8500/v1/agent/service/register?replace-existing-checks=1
注册完毕,通过 Consul Web 管理页面可以查看到两服务已注册成功。注意:这里需要启动 node-exporter 及 cadvisor-exporter,否则即使注册成功了,健康检测也不通过,在后边 Prometheus 上服务发现后状态也是不健康的,这里我就不在演示了,可以直接拿 Docker 启动二者即可。
接下来,我们修改 Prometheus 配置如下:
...
- job_name: 'consul-node-exporter'
consul_sd_configs:
- server: '172.30.12.100:8500'
services: []
relabel_configs:
- source_labels: [__meta_consul_tags]
regex: .*node-exporter.*
action: keep
- regex: __meta_consul_service_metadata_(.+)
action: labelmap
- job_name: 'consul-cadvisor-exproter'
consul_sd_configs:
- server: '172.30.12.100:8510'
services: []
- server: '172.30.12.100:8520'
services: []
relabel_configs:
- source_labels: [__meta_consul_tags]
regex: .*cadvisor-exporter.*
action: keep
- action: labelmap
regex: __meta_consul_service_metadata_(.+)
...
启动一下 Prometheus 服务,通过 web 界面看下 Target 中是否能够发现 Consul 中注册的服务。
可以看到,妥妥没有问题,这里 consul-node-exporter
我配置指向了 node1 Consul 服务地址,consul-cadvisor-exproter
配置指向了 node2、node3 Consul 服务,二者都能够正确发现之前注册的服务,因为 Consul 集群数据是保持同步的,无论连接哪一个节点,都能够获取到注册的服务信息,同理,我们也可以指定 consul_sd_configs
分别指向集群所有节点,这样即使某个节点挂掉,也不会影响 Prometheus 从 Consul 集群其他节点获取注册的服务,从而实现服务的高可用。
虽然我们可以将整个 Consul 集群 IP 添加到 Prometheus 的配置中,从而实现 Prometheus 从 Consul 集群获取注册的服务,实现服务的高可用,但是这里有个问题,如果 Consul 集群节点新增或者减少,那么 Prometheus 配置也得跟着修改了,这样不是很友好,我们可以在 Consul 集群前面使用 nginx 反向代理将请求负载均衡到后端 Consul 集群各节点服务上,这样 Prometheus 只需要配置代理地址即可,后期不需要更改了。
这里为了方便,nginx 也使用容器方式启动,新增配置文件 default.conf
如下:
upstream service_consul {
server 172.30.12.100:8500;
server 172.30.12.100:8510;
server 172.30.12.100:8520;
ip_hash;
}
server {
listen 80;
server_name 172.30.12.100;
index index.html index.htm;
#access_log /var/log/nginx/host.access.log main;
location / {
add_header Access-Control-Allow-Origin *;
proxy_next_upstream http_502 http_504 error timeout invalid_header;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_pass http://service_consul;
}
access_log /var/log/consul.access.log;
error_log /var/log/consul.error.log;
error_page 404 /404.html;
error_page 500 502 503 504 /50x.html;
location = /50x.html {
root /usr/share/nginx/html;
}
}
Docker 启动 nginx 服务,启动命令如下:
$ docker run --name nginx-consul -p 80:80 -v /root/prometheus/nginx/default.conf:/etc/nginx/conf.d/default.conf:ro -d nginx
这里就没啥好说的了,启动完成后,可以浏览器访问一下 http://172.30.12.100/ui
,妥妥也是没有问题的。
最后,我们修改一下 Prometheus 配置文件,将 server 指向 172.30.12.100
即可,修改配置如下:
...
- job_name: 'consul-node-exporter'
consul_sd_configs:
- server: '172.30.12.100'
services: []
relabel_configs:
- source_labels: [__meta_consul_tags]
regex: .*node-exporter.*
action: keep
- regex: __meta_consul_service_metadata_(.+)
action: labelmap
- job_name: 'consul-cadvisor-exproter'
consul_sd_configs:
- server: '172.30.12.100'
services: []
relabel_configs:
- source_labels: [__meta_consul_tags]
regex: .*cadvisor-exporter.*
action: keep
- action: labelmap
regex: __meta_consul_service_metadata_(.+)
...
重启 prometheus 服务,验证也是妥妥没问题的。