分布式理论 Raft

Raft 协议分析与实战：构建自己的分布式缓存

hashicorp/raft的应用实战与分析

Posted by pandaychen on April 15, 2022

0x00 前言

0x01 验证

按照如下组成一个raft集群：

raft节点 端口 peer端口 存储目录
Node-1 6000 7000 node0
Node-2 6001 7001 node1
Node-3 6002 7002 node2

1、启动集群，并成为leader

[root@VM_120_245_centos ~/stcache]# ./stcached --http=127.0.0.1:6000 --raft=127.0.0.1:7000 --node=./node0 -bootstrap=true


stCached: 2022/07/29 18:08:53 http server listen:127.0.0.1:6000
2022-07-29T18:08:53.158+0800 [INFO]  raft: initial configuration: index=0 servers=[]
2022-07-29T18:08:53.158+0800 [INFO]  raft: entering follower state: follower="Node at 127.0.0.1:7000 [Follower]" leader-address= leader-id=
2022-07-29T18:08:54.395+0800 [WARN]  raft: heartbeat timeout reached, starting election: last-leader-addr= last-leader-id=
2022-07-29T18:08:54.396+0800 [INFO]  raft: entering candidate state: node="Node at 127.0.0.1:7000 [Candidate]" term=2
2022-07-29T18:08:54.413+0800 [DEBUG] raft: votes: needed=1
2022-07-29T18:08:54.413+0800 [DEBUG] raft: vote granted: from=127.0.0.1:7000 term=2 tally=1
2022-07-29T18:08:54.413+0800 [INFO]  raft: election won: tally=1
2022-07-29T18:08:54.413+0800 [INFO]  raft: entering leader state: leader="Node at 127.0.0.1:7000 [Leader]"
stCached: 2022/07/29 18:08:54 become leader, enable write api #成为leader

2、启动第2个节点

[root@VM_120_245_centos ~/blog_backup/stcache/stcache]# ./stcached --http=127.0.0.1:6001 --raft=127.0.0.1:7001 --node=./node1 --join=127.0.0.1:6000
stCached: 2022/07/29 18:10:35 http server listen:127.0.0.1:6001
2022-07-29T18:10:35.580+0800 [INFO]  raft: initial configuration: index=0 servers=[]
2022-07-29T18:10:35.580+0800 [INFO]  raft: entering follower state: follower="Node at 127.0.0.1:7001 [Follower]" leader-address= leader-id=
2022-07-29T18:10:35.592+0800 [WARN]  raft: failed to get previous log: previous-index=3 last-index=0 error="log not found"
2022-07-29T18:11:08.417+0800 [ERROR] raft: failed to take snapshot: error="nothing new to snapshot"

3、启动第3个节点

[root@VM_120_245_centos ~/blog_backup/stcache/stcache]# ./stcached --http=127.0.0.1:6002 --raft=127.0.0.1:7002 --node=./node2 --join=127.0.0.1:6000
stCached: 2022/07/29 18:10:55 http server listen:127.0.0.1:6002
2022-07-29T18:10:55.231+0800 [INFO]  raft: initial configuration: index=0 servers=[]
2022-07-29T18:10:55.231+0800 [INFO]  raft: entering follower state: follower="Node at 127.0.0.1:7002 [Follower]" leader-address= leader-id=
2022-07-29T18:10:55.239+0800 [WARN]  raft: failed to get previous log: previous-index=4 last-index=0 error="log not found"

节点2、节点3加入集群后，leader日志： raft-1

4、写入/读取数据：

[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6000/set?key=abc&value=efg"
ok
[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6000/get?key=abc"
efg
[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6001/get?key=abc"
efg
[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6002/get?key=abc"
efg
[root@VM_120_245_centos ~]#

5、不可以向非leader写入数据

[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6001/set?key=abc&value=efg"
write method not allowed
[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6002/set?key=abc&value=efg"
write method not allowed

6、写入数据时leader/follower的状态变化

leader写入数据
raft-2

节点2（follower1）
raft-2

节点3（follower2）
raft-2

7、关闭leader节点
关闭节点1后，节点3通过选举成为了新的leader

raft-3

节点3可写：

[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6002/set?key=cba&value=gfe"
ok

8、查询follower变迁及数据

[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6001/get?key=cba"
gfe

9、将节点1重启，再加入集群，自动变为follower raft-3

leader状态： raft-3

再查询节点1的数据：

[root@VM_120_245_centos ~]# curl "http://127.0.0.1:6000/get?key=cba"
gfe

参考

CATALOG
    FEATURED TAGS
    Latex gRPC 负载均衡 OpenSSH Authentication Consul Etcd Kubernetes 性能优化 Python 分布式锁 WebConsole 后台开发 Golang OpenSource Nginx Vault 网络安全 Perl 分布式理论 Raft 正则表达式 Redis 分布式 限流 go-redis 微服务 反向代理 ReverseProxy Cache 缓存 连接池 OpenTracing GOMAXPROCS GoMicro 微服务框架 日志 zap Pool Kratos Hystrix 熔断 并发 Pipeline 证书 Prometheus Metrics PromQL Breaker 定时器 Timer Timeout Kafka Xorm MySQL Fasthttp bytebufferpool 任务队列 队列 异步队列 GOIM Pprof errgroup consistent-hash Zinx 网络框架 设计模式 HTTP Gateway Queue Docker 网关 Statefulset NFS Machinery Teleport Zero Trust Oxy 存储 Confd 热更新 OAuth SAML OpenID Openssl AES 微服务网关 IM KMS 安全 数据结构 hashtable Sort Asynq 基数树 Radix Crontab 热重启 系统编程 sarama Go-Zero RDP VNC 协程池 UDP hashmap 网络编程 自适应技术 环形队列 Ring Buffer Circular Buffer InnoDB timewheel GroupCache Jaeger GOSSIP CAP Bash websocket 事务 GC TLS singleflight 闭包 Helm network iptables MITM HTTPS Tap Tun 路由 wireguard gvisor Git NAT 协议栈 Envoy FRP DPI gopacket Cgroup Namespace DNS eBPF GoZero Gost Clash Tracee gopsutil Linux HIDS ELKEID XDP TC Systemd netlink Kernel BCC rootkit bpftrace
    FRIENDS