/
worker.go
218 lines (187 loc) · 5.06 KB
/
worker.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
// Copyright 2022 Canonical Ltd.
// Licensed under the AGPLv3, see LICENCE file for details.
package state
import (
"math/rand"
"sync"
"time"
"github.com/juju/errors"
"github.com/juju/worker/v3"
"github.com/juju/worker/v3/catacomb"
"gopkg.in/tomb.v2"
"github.com/juju/juju/state"
"github.com/juju/juju/wrench"
)
type stateWorker struct {
catacomb catacomb.Catacomb
stTracker StateTracker
pingInterval time.Duration
setStatePool func(*state.StatePool)
cleanupOnce sync.Once
pool *state.StatePool
modelStateWorkers map[string]worker.Worker
}
func (w *stateWorker) loop() error {
var err error
w.pool, err = w.stTracker.Use()
if err != nil {
return errors.Trace(err)
}
defer func() { _ = w.stTracker.Done() }()
systemState, err := w.pool.SystemState()
if err != nil {
return errors.Trace(err)
}
w.setStatePool(w.pool)
defer w.setStatePool(nil)
modelWatcher := systemState.WatchModelLives()
_ = w.catacomb.Add(modelWatcher)
wrenchTimer := time.NewTimer(30 * time.Second)
defer wrenchTimer.Stop()
w.modelStateWorkers = make(map[string]worker.Worker)
for {
select {
case <-w.catacomb.Dying():
return w.catacomb.ErrDying()
case modelUUIDs := <-modelWatcher.Changes():
for _, modelUUID := range modelUUIDs {
if err := w.processModelLifeChange(modelUUID); err != nil {
return errors.Trace(err)
}
}
// Useful for tracking down some bugs that occur when
// mongo is overloaded.
case <-wrenchTimer.C:
if wrench.IsActive("state-worker", "io-timeout") {
return errors.Errorf("wrench simulating i/o timeout!")
}
wrenchTimer.Reset(30 * time.Second)
}
}
}
// Report conforms to the Dependency Engine Report() interface, giving an opportunity to introspect
// what is going on at runtime.
func (w *stateWorker) Report() map[string]interface{} {
return w.stTracker.Report()
}
func (w *stateWorker) processModelLifeChange(modelUUID string) error {
model, hp, err := w.pool.GetModel(modelUUID)
if err != nil {
if errors.IsNotFound(err) {
// Model has been removed from state.
logger.Debugf("model %q removed from state", modelUUID)
w.remove(modelUUID)
return nil
}
return errors.Trace(err)
}
defer hp.Release()
if model.Life() == state.Dead {
// Model is Dead, and will soon be removed from state.
logger.Debugf("model %q is dead", modelUUID)
w.remove(modelUUID)
return nil
}
if w.modelStateWorkers[modelUUID] == nil {
mw := newModelStateWorker(w.pool, modelUUID, w.pingInterval)
w.modelStateWorkers[modelUUID] = mw
_ = w.catacomb.Add(mw)
}
return nil
}
func (w *stateWorker) remove(modelUUID string) {
if worker, ok := w.modelStateWorkers[modelUUID]; ok {
worker.Kill()
delete(w.modelStateWorkers, modelUUID)
}
_, _ = w.pool.Remove(modelUUID)
}
// Kill is part of the worker.Worker interface.
func (w *stateWorker) Kill() {
w.catacomb.Kill(nil)
}
// Wait is part of the worker.Worker interface.
func (w *stateWorker) Wait() error {
err := w.catacomb.Wait()
w.cleanupOnce.Do(func() {
// Make sure the worker has exited before closing state.
if err := w.stTracker.Done(); err != nil {
logger.Warningf("error releasing state: %v", err)
}
})
return err
}
type modelStateWorker struct {
tomb tomb.Tomb
pool *state.StatePool
modelUUID string
pingInterval time.Duration
}
func newModelStateWorker(
pool *state.StatePool,
modelUUID string,
pingInterval time.Duration,
) worker.Worker {
w := &modelStateWorker{
pool: pool,
modelUUID: modelUUID,
pingInterval: pingInterval,
}
w.tomb.Go(w.loop)
return w
}
func (w *modelStateWorker) loop() error {
st, err := w.pool.Get(w.modelUUID)
if err != nil {
if errors.IsNotFound(err) {
// Ignore not found error here, because the pooledState has already
// been removed.
return nil
}
return errors.Trace(err)
}
defer func() {
st.Release()
_, _ = w.pool.Remove(w.modelUUID)
}()
// Jitter the interval so that each model doesn't attempt to connect to
// mongo at potentially the same time.
interval := w.pingInterval + jitter(time.Millisecond*200)
// If the state ping fails, attempt to retry the ping, before returning.
var pingErr error
for attempt := 0; attempt < 5; {
select {
case <-w.tomb.Dying():
return tomb.ErrDying
case <-time.After(interval):
if pingErr = st.Ping(); pingErr != nil {
// Reduce the next ping interval to fail early in case mongo
// has actually died. This should prevent the worst case
// scenario of a large initial ping interval.
interval = maxDuration(interval/2, time.Second)
attempt++
continue
}
interval = w.pingInterval
attempt = 0
}
}
return errors.Annotate(pingErr, "state ping failed")
}
// Kill is part of the worker.Worker interface.
func (w *modelStateWorker) Kill() {
w.tomb.Kill(nil)
}
// Wait is part of the worker.Worker interface.
func (w *modelStateWorker) Wait() error {
return w.tomb.Wait()
}
func maxDuration(a, b time.Duration) time.Duration {
if a > b {
return a
}
return b
}
func jitter(amount time.Duration) time.Duration {
return time.Duration((rand.Float64() - 0.5) * float64(amount))
}