├── README.md # 操作指南
├── main.go # demo示例
└── redis
└── redis.go # redis封装库（长连接与读写分离）
redis安装：see https://github.com/caijinlin/centos-dockerfile/blob/master/soft/redis/build.sh
redis启动: see https://github.com/caijinlin/centos-dockerfile/blob/master/soft/redis/control.sh
执行命令后会启动两个redis server, master指向6379端口，slave指向6380端口，三个sentinel，监控redis master
登录master
bin/redis-cli -p 6379
登录slave
bin/redis-cli -p 6380
查看连接clients及设置
bin/redis-cli -p 6379 info clients
bin/redis-cli -p 6379 client list
bin/redis-cli -p 6379 config get maxclients
登录各个sentinel
bin/redis-cli -p 26379
bin/redis-cli -p 26380
bin/redis-cli -p 26381
info 查看监听的redis master实例
sentinel_masters:1
sentinel_tilt:0
sentinel_running_scripts:0
sentinel_scripts_queue_length:0
sentinel_simulate_failure_flags:0
master0:name=localmaster,status=ok,address=127.0.0.1:6379,slaves=1,sentinels=3
关闭主数据库
bin/redis-cli -p 6379 shutdown
查看sentinel日志
tail -f log/redis-sentinel-*
追加了如下内容
=> log/redis-sentinel-26381.log <==
24331:X 27 Mar 2019 18:33:55.121 # +promoted-slave slave 127.0.0.1:6380 127.0.0.1 6380 @ localmaster 127.0.0.1 6379
24331:X 27 Mar 2019 18:33:55.121 # +failover-state-reconf-slaves master localmaster 127.0.0.1 6379
24331:X 27 Mar 2019 18:33:55.191 * +slave-reconf-sent slave 172.18.0.1:6380 172.18.0.1 6380 @ localmaster 127.0.0.1 6379
24331:X 27 Mar 2019 18:33:55.191 # +failover-end master localmaster 127.0.0.1 6379
24331:X 27 Mar 2019 18:33:55.193 # +switch-master localmaster 127.0.0.1 6379 127.0.0.1 6380
24331:X 27 Mar 2019 18:33:55.193 * +slave slave 172.18.0.1:6380 172.18.0.1 6380 @ localmaster 127.0.0.1 6380
24331:X 27 Mar 2019 18:33:55.193 * +slave slave 127.0.0.1:6379 127.0.0.1 6379 @ localmaster 127.0.0.1 6380
==> log/redis-sentinel-26379.log <==
24327:X 27 Mar 2019 18:33:55.194 # +config-update-from sentinel b7d901f1775f63057bad25fdd8117f008ba25532 127.0.0.1 26381 @ localmaster 127.0.0.1 6379
24327:X 27 Mar 2019 18:33:55.195 # +switch-master localmaster 127.0.0.1 6379 127.0.0.1 6380
==> log/redis-sentinel-26380.log <==
24329:X 27 Mar 2019 18:33:55.194 # +config-update-from sentinel b7d901f1775f63057bad25fdd8117f008ba25532 127.0.0.1 26381 @ localmaster 127.0.0.1 6379
24329:X 27 Mar 2019 18:33:55.195 # +switch-master localmaster 127.0.0.1 6379 127.0.0.1 6380
启动故障数据库
bin/redis-server ./conf/redis_6379.conf
查看sentinel日志，故障数据库加入slaves了
==> log/redis-sentinel-26381.log <==
24331:X 28 Mar 2019 10:36:28.922 * +convert-to-slave slave 127.0.0.1:6379 127.0.0.1 6379 @ localmaster 127.0.0.1 6380
- 辅助库
https://github.com/gomodule/redigo
- 运行demo
go get github.com/gomodule/redigo
go run main.go
- 测试用例
# 模拟20个请求，10个并发 curl 请求
ab -c 10 -n 20 http://127.0.0.1:3333/
- 查看连接数
netstat -anp | grep 6380
- 核心代码解析
// get prunes stale connections and returns a connection from the idle list or
// creates a new connection.
func (p *Pool) get(ctx context.Context) (*poolConn, error) {
// Handle limit for p.Wait == true.
var waited time.Duration
if p.Wait && p.MaxActive > 0 {
p.lazyInit()
// wait indicates if we believe it will block so its not 100% accurate
// however for stats it should be good enough.
wait := len(p.ch) == 0
var start time.Time
if wait {
start = time.Now()
}
if ctx == nil {
<-p.ch
} else {
select {
case <-p.ch:
case <-ctx.Done():
return nil, ctx.Err()
}
}
if wait {
waited = time.Since(start)
}
}
// 防止高并发时两个请求用一个socket导致错误，加锁获取连接
p.mu.Lock()
if waited > 0 {
p.waitCount++
p.waitDuration += waited
}
// 移除过了IdleTimeout时间没有使用的连接
// Prune stale connections at the back of the idle list.
if p.IdleTimeout > 0 {
n := p.idle.count
for i := 0; i < n && p.idle.back != nil && p.idle.back.t.Add(p.IdleTimeout).Before(nowFunc()); i++ {
pc := p.idle.back
p.idle.popBack()
p.mu.Unlock()
pc.c.Close()
p.mu.Lock()
p.active--
}
}
// 从最大长度为MaxIdle的idlelist弹出一个连接
// Get idle connection from the front of idle list.
for p.idle.front != nil {
pc := p.idle.front
p.idle.popFront()
p.mu.Unlock()
if (p.TestOnBorrow == nil || p.TestOnBorrow(pc.c, pc.t) == nil) &&
(p.MaxConnLifetime == 0 || nowFunc().Sub(pc.created) < p.MaxConnLifetime) {
return pc, nil
}
pc.c.Close()
p.mu.Lock()
p.active--
}
// 异常连接报错
// Check for pool closed before dialing a new connection.
if p.closed {
p.mu.Unlock()
return nil, errors.New("redigo: get on closed pool")
}
// 该时刻超出设置的活跃连接数报错
// Handle limit for p.Wait == false.
if !p.Wait && p.MaxActive > 0 && p.active >= p.MaxActive {
p.mu.Unlock()
return nil, ErrPoolExhausted
}
p.active++
p.mu.Unlock()
c, err := p.dial(ctx)
if err != nil {
c = nil
p.mu.Lock()
p.active--
if p.ch != nil && !p.closed {
p.ch <- struct{}{}
}
p.mu.Unlock()
}
return &poolConn{c: c, created: nowFunc()}, err
}