-
Notifications
You must be signed in to change notification settings - Fork 2
/
chash.go
308 lines (274 loc) · 7.11 KB
/
chash.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
305
306
307
308
package chash
import (
"errors"
"fmt"
"golang.org/x/exp/slices"
"sort"
"sync"
"github.com/cespare/xxhash"
)
var (
ErrMemberExists = errors.New("member exists")
ErrMemberNotExists = errors.New("member not exists")
ErrPartitionNotExists = errors.New("partition not exists")
ErrInvalidCapacity = errors.New("member capacity must be > 0")
)
type defaultHasher struct{}
func (h defaultHasher) Sum64(data []byte) uint64 {
return xxhash.Sum64(data)
}
func New(c Config) (CHash, error) {
if c.Hasher == nil {
c.Hasher = defaultHasher{}
}
if c.ReplicationFactor == 0 {
c.ReplicationFactor = 1
}
if c.MultiplyFactor <= 0 {
c.MultiplyFactor = defaultMultiplyFactor
}
h := &cHash{config: c}
if err := h.init(); err != nil {
return nil, err
}
return h, nil
}
type CHash interface {
// AddMembers adds one or more members to the cluster
// May return ErrInvalidCapacity if member capacity less or equal 0
// May return ErrMemberExists if member was added before
AddMembers(members ...Member) error
// RemoveMembers removes members with given ids
RemoveMembers(memberIds ...string) error
// Reconfigure replaces all members list
Reconfigure(members []Member) error
// GetMembers returns list of members for given key
// Members count will be equal replication factor or total members count (if it is less than the replication factor)
GetMembers(key string) []Member
// GetPartition returns partition number for given key
GetPartition(key string) int
// GetPartitionMembers return members by partition number
GetPartitionMembers(partId int) ([]Member, error)
// Distribute members by partitions
// Must be called if you changed members' capacity
Distribute()
// PartitionCount returns configured partitions count
PartitionCount() int
}
type Member interface {
Id() string
Capacity() float64
}
type Hasher interface {
Sum64([]byte) uint64
}
type Config struct {
// Hasher implementation (optional), by default, will be used xxhash
Hasher Hasher
// PartitionCount - how many virtual partitions will be distributed by nodes, reasonable values from 100 to 10k
PartitionCount uint64
// ReplicationFactor - how many nodes expected for GetMembers
ReplicationFactor int
// Multiply Factor (optional) - this value multiplied for member capacity means how many times a member will be added to the hash ring. The default value is 2000.
MultiplyFactor int
}
func (c Config) Validate() (err error) {
if c.ReplicationFactor < 1 {
return fmt.Errorf("replcation factor must be great or equal 1")
}
if c.PartitionCount < 10 {
return fmt.Errorf("patiotin count must be great ir qual 10")
}
return
}
const defaultMultiplyFactor = 2000
type cHash struct {
config Config
members map[string]Member
membersSet members
piecesPerMember map[string]int
partitions [][]Member
partitionHashes []uint64
mu sync.RWMutex
}
func (c *cHash) init() (err error) {
if err = c.config.Validate(); err != nil {
return
}
c.members = make(map[string]Member)
c.partitionHashes = make([]uint64, c.config.PartitionCount)
c.partitions = make([][]Member, c.config.PartitionCount)
for i := range c.partitionHashes {
c.partitionHashes[i] = c.config.Hasher.Sum64([]byte(fmt.Sprint("p", i)))
}
return
}
func (c *cHash) AddMembers(members ...Member) error {
c.mu.Lock()
defer c.mu.Unlock()
for _, m := range members {
if m.Capacity() <= 0 {
return ErrInvalidCapacity
}
if _, ok := c.members[m.Id()]; ok {
return ErrMemberExists
}
}
return c.addMembers(members...)
}
func (c *cHash) addMembers(members ...Member) error {
for _, m := range members {
// generating enough virtual members for better hash distribution
for i := 0; i < int(float64(c.config.MultiplyFactor)*m.Capacity()); i++ {
c.membersSet = append(c.membersSet, member{
hash: c.config.Hasher.Sum64([]byte(fmt.Sprint(m.Id(), i))),
Member: m,
})
}
c.members[m.Id()] = m
}
sort.Sort(c.membersSet)
c.distribute()
return nil
}
func (c *cHash) RemoveMembers(memberIds ...string) error {
c.mu.Lock()
defer c.mu.Unlock()
for _, mId := range memberIds {
if _, ok := c.members[mId]; !ok {
return ErrMemberNotExists
}
}
discard := func(ids ...string) members {
idx := 0
for _, el := range c.membersSet {
if !slices.Contains(ids, el.Id()) {
c.membersSet[idx] = el
idx++
}
}
return c.membersSet[:idx]
}
c.membersSet = discard(memberIds...)
for _, mId := range memberIds {
delete(c.members, mId)
}
c.distribute()
return nil
}
func (c *cHash) Reconfigure(members []Member) error {
c.mu.Lock()
defer c.mu.Unlock()
for _, m := range members {
if m.Capacity() <= 0 {
return ErrInvalidCapacity
}
}
c.members = make(map[string]Member)
c.membersSet = c.membersSet[:0]
return c.addMembers(members...)
}
func (c *cHash) GetMembers(key string) []Member {
c.mu.RLock()
defer c.mu.RUnlock()
return c.partitions[c.getPartition(key)]
}
func (c *cHash) GetPartition(key string) int {
c.mu.RLock()
defer c.mu.RUnlock()
return c.getPartition(key)
}
func (c *cHash) PartitionCount() int {
return int(c.config.PartitionCount)
}
func (c *cHash) getPartition(key string) int {
h := c.config.Hasher.Sum64([]byte(key))
return int(h % c.config.PartitionCount)
}
func (c *cHash) GetPartitionMembers(partId int) ([]Member, error) {
c.mu.RLock()
defer c.mu.RUnlock()
if partId < 0 || partId >= int(c.config.PartitionCount) {
return nil, ErrPartitionNotExists
}
return c.partitions[partId], nil
}
func (c *cHash) Distribute() {
c.mu.Lock()
defer c.mu.Unlock()
c.distribute()
}
func (c *cHash) distribute() {
if len(c.membersSet) == 0 {
for i := range c.partitions {
c.partitions[i] = nil
}
return
}
var totalCapacity float64
rf := c.config.ReplicationFactor
if len(c.members) < rf {
rf = len(c.members)
}
for _, m := range c.members {
totalCapacity += m.Capacity()
}
c.piecesPerMember = map[string]int{}
for _, m := range c.members {
p := int((float64(c.config.PartitionCount)*float64(rf))/(totalCapacity/m.Capacity())) + 1
c.piecesPerMember[m.Id()] = p
}
var buf = make([]string, rf)
for i, h := range c.partitionHashes {
if len(c.partitions[i]) != rf {
c.partitions[i] = make([]Member, rf)
}
c.fillClosest(c.membersSet, h, c.partitions[i], buf)
}
}
func (c *cHash) fillClosest(m members, h uint64, ms []Member, buf []string) {
idx := sort.Search(len(m), func(i int) bool {
return m[i].hash >= h
})
var found int
var maxOverflow int
var foundId = buf[:0]
var isAlreadyFound = func(id string) bool {
return slices.Contains(foundId, id)
}
for found < len(ms) {
if idx == m.Len() {
idx = 0
}
if isAlreadyFound(m[idx].Id()) {
maxOverflow++
idx++
continue
}
if c.piecesPerMember[m[idx].Id()] > -maxOverflow {
c.piecesPerMember[m[idx].Id()]--
ms[found] = m[idx].Member
foundId = append(foundId, m[idx].Id())
found++
}
idx++
}
}
type member struct {
hash uint64
Member
}
type members []member
func (m members) Len() int {
return len(m)
}
func (m members) Less(i, j int) bool {
if m[i].hash == m[j].hash {
return m[i].Id() < m[j].Id()
} else {
return m[i].hash < m[j].hash
}
}
func (m members) Swap(i, j int) {
m[i], m[j] = m[j], m[i]
}