-
Notifications
You must be signed in to change notification settings - Fork 1
/
evict_leader.go
254 lines (225 loc) · 7.04 KB
/
evict_leader.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
package schedulers
import (
"errors"
"strconv"
"sync"
"github.com/deepfabric/prophet/core"
"github.com/deepfabric/prophet/schedule"
"github.com/deepfabric/prophet/schedule/filter"
"github.com/deepfabric/prophet/schedule/operator"
"github.com/deepfabric/prophet/schedule/opt"
"github.com/deepfabric/prophet/storage"
"github.com/deepfabric/prophet/util"
)
const (
// EvictLeaderName is evict leader scheduler name.
EvictLeaderName = "evict-leader-scheduler"
// EvictLeaderType is evict leader scheduler type.
EvictLeaderType = "evict-leader"
// EvictLeaderBatchSize is the number of operators to to transfer
// leaders by one scheduling
EvictLeaderBatchSize = 3
lastContainerDeleteInfo = "The last container has been deleted"
)
func init() {
schedule.RegisterSliceDecoderBuilder(EvictLeaderType, func(args []string) schedule.ConfigDecoder {
return func(v interface{}) error {
if len(args) != 1 {
return errors.New("scheduler error configuration")
}
conf, ok := v.(*evictLeaderSchedulerConfig)
if !ok {
return errors.New("scheduler not found")
}
id, err := strconv.ParseUint(args[0], 10, 64)
if err != nil {
return errors.New("scheduler error configuration")
}
ranges, err := getKeyRanges(args[1:])
if err != nil {
return err
}
conf.ContainerIDWithRanges[id] = ranges
return nil
}
})
schedule.RegisterScheduler(EvictLeaderType, func(opController *schedule.OperatorController, storage storage.Storage, decoder schedule.ConfigDecoder) (schedule.Scheduler, error) {
conf := &evictLeaderSchedulerConfig{ContainerIDWithRanges: make(map[uint64][]core.KeyRange), storage: storage}
if err := decoder(conf); err != nil {
return nil, err
}
conf.cluster = opController.GetCluster()
return newEvictLeaderScheduler(opController, conf), nil
})
}
type evictLeaderSchedulerConfig struct {
mu sync.RWMutex
storage storage.Storage
ContainerIDWithRanges map[uint64][]core.KeyRange `json:"container-id-ranges"`
cluster opt.Cluster
}
func (conf *evictLeaderSchedulerConfig) BuildWithArgs(args []string) error {
if len(args) != 1 {
return errors.New("scheduler error coniguration")
}
id, err := strconv.ParseUint(args[0], 10, 64)
if err != nil {
return errors.New("scheduler error coniguration")
}
ranges, err := getKeyRanges(args[1:])
if err != nil {
return err
}
conf.mu.Lock()
defer conf.mu.Unlock()
conf.ContainerIDWithRanges[id] = ranges
return nil
}
func (conf *evictLeaderSchedulerConfig) Clone() *evictLeaderSchedulerConfig {
conf.mu.RLock()
defer conf.mu.RUnlock()
return &evictLeaderSchedulerConfig{
ContainerIDWithRanges: conf.ContainerIDWithRanges,
}
}
func (conf *evictLeaderSchedulerConfig) Persist() error {
name := conf.getSchedulerName()
conf.mu.RLock()
defer conf.mu.RUnlock()
data, err := schedule.EncodeConfig(conf)
if err != nil {
return err
}
return conf.storage.SaveScheduleConfig(name, data)
}
func (conf *evictLeaderSchedulerConfig) getSchedulerName() string {
return EvictLeaderName
}
func (conf *evictLeaderSchedulerConfig) getRanges(id uint64) []string {
conf.mu.RLock()
defer conf.mu.RUnlock()
var res []string
ranges := conf.ContainerIDWithRanges[id]
for index := range ranges {
res = append(res, (string)(ranges[index].StartKey), (string)(ranges[index].EndKey))
}
return res
}
func (conf *evictLeaderSchedulerConfig) mayBeRemoveContainerFromConfig(id uint64) (succ bool, last bool) {
conf.mu.Lock()
defer conf.mu.Unlock()
_, exists := conf.ContainerIDWithRanges[id]
succ, last = false, false
if exists {
delete(conf.ContainerIDWithRanges, id)
conf.cluster.ResumeLeaderTransfer(id)
succ = true
last = len(conf.ContainerIDWithRanges) == 0
}
return succ, last
}
type evictLeaderScheduler struct {
*BaseScheduler
conf *evictLeaderSchedulerConfig
}
// newEvictLeaderScheduler creates an admin scheduler that transfers all leaders
// out of a container.
func newEvictLeaderScheduler(opController *schedule.OperatorController, conf *evictLeaderSchedulerConfig) schedule.Scheduler {
base := NewBaseScheduler(opController)
return &evictLeaderScheduler{
BaseScheduler: base,
conf: conf,
}
}
func (s *evictLeaderScheduler) GetName() string {
return EvictLeaderName
}
func (s *evictLeaderScheduler) GetType() string {
return EvictLeaderType
}
func (s *evictLeaderScheduler) EncodeConfig() ([]byte, error) {
s.conf.mu.RLock()
defer s.conf.mu.RUnlock()
return schedule.EncodeConfig(s.conf)
}
func (s *evictLeaderScheduler) Prepare(cluster opt.Cluster) error {
s.conf.mu.RLock()
defer s.conf.mu.RUnlock()
var res error
for id := range s.conf.ContainerIDWithRanges {
if err := cluster.PauseLeaderTransfer(id); err != nil {
res = err
}
}
return res
}
func (s *evictLeaderScheduler) Cleanup(cluster opt.Cluster) {
s.conf.mu.RLock()
defer s.conf.mu.RUnlock()
for id := range s.conf.ContainerIDWithRanges {
cluster.ResumeLeaderTransfer(id)
}
}
func (s *evictLeaderScheduler) IsScheduleAllowed(cluster opt.Cluster) bool {
return s.OpController.OperatorCount(operator.OpLeader) < cluster.GetOpts().GetLeaderScheduleLimit()
}
func (s *evictLeaderScheduler) scheduleOnce(cluster opt.Cluster) []*operator.Operator {
var ops []*operator.Operator
for id, ranges := range s.conf.ContainerIDWithRanges {
res := cluster.RandLeaderResource(id, ranges, opt.HealthResource(cluster))
if res == nil {
schedulerCounter.WithLabelValues(s.GetName(), "no-leader").Inc()
continue
}
target := filter.NewCandidates(cluster.GetFollowerContainers(res)).
FilterTarget(cluster.GetOpts(), &filter.ContainerStateFilter{ActionScope: EvictLeaderName, TransferLeader: true}).
RandomPick()
if target == nil {
schedulerCounter.WithLabelValues(s.GetName(), "no-target-container").Inc()
continue
}
op, err := operator.CreateTransferLeaderOperator(EvictLeaderType, cluster, res, res.GetLeader().GetContainerID(), target.Meta.ID(), operator.OpLeader)
if err != nil {
util.GetLogger().Debugf("create evict leader operator failed with %+v",
err)
continue
}
op.SetPriorityLevel(core.HighPriority)
op.Counters = append(op.Counters, schedulerCounter.WithLabelValues(s.GetName(), "new-operator"))
ops = append(ops, op)
}
return ops
}
func (s *evictLeaderScheduler) uniqueAppend(dst []*operator.Operator, src ...*operator.Operator) []*operator.Operator {
resIDs := make(map[uint64]struct{})
for i := range dst {
resIDs[dst[i].ResourceID()] = struct{}{}
}
for i := range src {
if _, ok := resIDs[src[i].ResourceID()]; ok {
continue
}
resIDs[src[i].ResourceID()] = struct{}{}
dst = append(dst, src[i])
}
return dst
}
func (s *evictLeaderScheduler) Schedule(cluster opt.Cluster) []*operator.Operator {
schedulerCounter.WithLabelValues(s.GetName(), "schedule").Inc()
var ops []*operator.Operator
s.conf.mu.RLock()
defer s.conf.mu.RUnlock()
for i := 0; i < EvictLeaderBatchSize; i++ {
once := s.scheduleOnce(cluster)
// no more resources
if len(once) == 0 {
break
}
ops = s.uniqueAppend(ops, once...)
// the batch has been fulfilled
if len(ops) > EvictLeaderBatchSize {
break
}
}
return ops
}