-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathgroup.go
304 lines (244 loc) · 6.13 KB
/
group.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
package sentinel
import (
"errors"
"fmt"
"net"
"strconv"
"strings"
"sync"
"time"
"github.com/garyburd/redigo/redis"
)
var (
logger = newLogger()
)
var (
ErrNodeDown = errors.New("Node is down")
ErrNodeAlive = errors.New("Node may be still alive")
ErrNoCandidate = errors.New("no proper candidate to be promoted to master")
ErrNodeType = errors.New("Node is not the expected type")
)
const (
MasterType = "master"
SlaveType = "slave"
)
const (
ConnectState = "connect"
ConnectingState = "connecting"
ConnectedState = "connected"
SyncState = "sync"
)
// A node represents a real redis server
type Node struct {
// Redis address, only support tcp now
Addr string
// Replication offset
Offset int64
conn redis.Conn
}
func (n *Node) String() string {
return fmt.Sprintf("{addr: %s offset: %d}", n.Addr, n.Offset)
}
func (n *Node) doCommand(cmd string, args ...interface{}) (interface{}, error) {
var err error
var v interface{}
for i := 0; i < 3; i++ {
if n.conn == nil {
n.conn, err = redis.DialTimeout("tcp", n.Addr, 5*time.Second, 0, 0)
if err != nil {
logger.Errorf("dial %s error: %v, try again", n.Addr, err)
continue
}
}
v, err = n.conn.Do(cmd, args...)
if err != nil {
logger.Errorf("do %s command for %s error: %v, try again", cmd, n.Addr, err)
n.conn.Close()
n.conn = nil
continue
} else {
return v, nil
}
}
// go here means do command error, maybe redis is down.
return nil, err
}
func (n *Node) doRole() ([]interface{}, error) {
return redis.Values(n.doCommand("ROLE"))
}
func (n *Node) ping() error {
_, err := n.doCommand("PING")
return err
}
func (n *Node) slaveof(host string, port string) error {
_, err := n.doCommand("SLAVEOF", host, port)
return err
}
func (n *Node) doRelpInfo() (map[string]string, error) {
v, err := redis.String(n.doCommand("INFO", "REPLICATION"))
if err != nil {
return nil, err
}
seps := strings.Split(v, "\r\n")
// skip first line, is # Replication
seps = seps[1:]
m := make(map[string]string, len(seps))
for _, s := range seps {
kv := strings.SplitN(s, ":", 2)
if len(kv) == 2 {
m[strings.TrimSpace(kv[0])] = strings.TrimSpace(kv[1])
}
}
return m, nil
}
func (n *Node) close() {
if n.conn != nil {
n.conn.Close()
n.conn = nil
}
}
// A group contains a Redis master and one or more slaves
// It will use role command per second to check master's alive
// and find slaves automatically.
type Group struct {
Master *Node
Slaves map[string]*Node
CheckErrNum int32
m sync.Mutex
}
func newGroup(masterAddr string) *Group {
g := new(Group)
g.Master = &Node{Addr: masterAddr}
g.Slaves = make(map[string]*Node)
return g
}
func (g *Group) Close() {
g.m.Lock()
defer g.m.Unlock()
g.Master.close()
for _, slave := range g.Slaves {
slave.close()
}
}
func (g *Group) Check() error {
g.m.Lock()
defer g.m.Unlock()
err := g.doRole()
if err != nil {
g.CheckErrNum += 1
} else {
g.CheckErrNum = 0
}
return err
}
func (g *Group) doRole() error {
v, err := g.Master.doRole()
if err != nil {
return ErrNodeDown
}
// the first line is server type
serverType, _ := redis.String(v[0], nil)
if serverType != MasterType {
logger.Errorf("server %s is not master now", g.Master.Addr)
return ErrNodeType
}
// second is master replication offset,
g.Master.Offset, _ = redis.Int64(v[1], nil)
// then slave list [host, port, offset]
slaves, _ := redis.Values(v[2], nil)
nodes := make(map[string]*Node, len(slaves))
for i := 0; i < len(slaves); i++ {
ss, _ := redis.Strings(slaves[i], nil)
var n Node
n.Addr = fmt.Sprintf("%s:%s", ss[0], ss[1])
n.Offset, _ = strconv.ParseInt(fmt.Sprintf("%s", ss[2]), 10, 64)
nodes[n.Addr] = &n
}
// we don't care slave add or remove too much, so only log
for addr := range nodes {
if _, ok := g.Slaves[addr]; !ok {
logger.Infof("slave %s added", addr)
}
}
for addr, slave := range g.Slaves {
if _, ok := nodes[addr]; !ok {
logger.Infof("slave %s removed", addr)
slave.close()
}
}
g.Slaves = nodes
return nil
}
func (g *Group) Ping() error {
g.m.Lock()
defer g.m.Unlock()
return g.Master.ping()
}
// Elect a best slave which has the most up-to-date data with master
func (g *Group) Elect() (string, error) {
g.m.Lock()
defer g.m.Unlock()
var addr string
var checkOffset int64 = 0
var checkPriority int = 0
for _, slave := range g.Slaves {
m, err := slave.doRelpInfo()
if err != nil {
logger.Infof("slave %s get replication info err %v, skip it", slave.Addr, err)
continue
}
if m["slave"] == MasterType {
logger.Errorf("server %s is not slave now, skip it", slave.Addr)
continue
}
if m["master_link_status"] == "up" {
logger.Infof("slave %s master_link_status is up, master %s may be not down???",
slave.Addr, g.Master.Addr)
return "", ErrNodeAlive
}
priority, _ := strconv.Atoi(m["slave_priority"])
replOffset, _ := strconv.ParseInt(m["slave_repl_offset"], 10, 64)
used := false
// like redis-sentinel, first check priority, then salve repl offset
if checkPriority < priority {
used = true
} else if checkPriority == priority {
if checkOffset < replOffset {
used = true
}
}
if used {
addr = slave.Addr
checkPriority = priority
checkOffset = replOffset
}
}
if len(addr) == 0 {
logger.Errorf("no proper candidate to be promoted")
return "", ErrNoCandidate
}
logger.Infof("select slave %s as new master, priority:%d, repl_offset:%d", addr, checkPriority, checkOffset)
return addr, nil
}
// Promote the slave to master, then let other slaves replicate from it
func (g *Group) Promote(addr string) error {
g.m.Lock()
defer g.m.Unlock()
node := g.Slaves[addr]
if err := node.slaveof("no", "one"); err != nil {
return err
}
delete(g.Slaves, addr)
g.Master = node
host, port, _ := net.SplitHostPort(addr)
for _, slave := range g.Slaves {
if err := slave.slaveof(host, port); err != nil {
// if we go here, the replication topology may be wrong
// so use fatal level and we should fix it manually
logger.Fatalf("slaveof %s to master %s err %v", slave.Addr, addr, err)
} else {
logger.Infof("slaveof %s to master %s ok", slave.Addr, addr)
}
}
return nil
}