/
compactor.go
360 lines (300 loc) · 8.58 KB
/
compactor.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
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
package bgs
import (
"context"
"fmt"
"sync"
"time"
"github.com/bluesky-social/indigo/carstore"
"github.com/bluesky-social/indigo/models"
"go.opentelemetry.io/otel"
"go.opentelemetry.io/otel/attribute"
)
type queueItem struct {
uid models.Uid
fast bool
}
// uniQueue is a queue that only allows one instance of a given uid
type uniQueue struct {
q []queueItem
members map[models.Uid]struct{}
lk sync.Mutex
}
// Append appends a uid to the end of the queue if it doesn't already exist
func (q *uniQueue) Append(uid models.Uid, fast bool) {
q.lk.Lock()
defer q.lk.Unlock()
if _, ok := q.members[uid]; ok {
return
}
q.q = append(q.q, queueItem{uid: uid, fast: fast})
q.members[uid] = struct{}{}
compactionQueueDepth.Inc()
}
// Prepend prepends a uid to the beginning of the queue if it doesn't already exist
func (q *uniQueue) Prepend(uid models.Uid, fast bool) {
q.lk.Lock()
defer q.lk.Unlock()
if _, ok := q.members[uid]; ok {
return
}
q.q = append([]queueItem{{uid: uid, fast: fast}}, q.q...)
q.members[uid] = struct{}{}
compactionQueueDepth.Inc()
}
// Has returns true if the queue contains the given uid
func (q *uniQueue) Has(uid models.Uid) bool {
q.lk.Lock()
defer q.lk.Unlock()
_, ok := q.members[uid]
return ok
}
// Remove removes the given uid from the queue
func (q *uniQueue) Remove(uid models.Uid) {
q.lk.Lock()
defer q.lk.Unlock()
if _, ok := q.members[uid]; !ok {
return
}
for i, item := range q.q {
if item.uid == uid {
q.q = append(q.q[:i], q.q[i+1:]...)
break
}
}
delete(q.members, uid)
compactionQueueDepth.Dec()
}
// Pop pops the first item off the front of the queue
func (q *uniQueue) Pop() (*queueItem, bool) {
q.lk.Lock()
defer q.lk.Unlock()
if len(q.q) == 0 {
return nil, false
}
item := q.q[0]
q.q = q.q[1:]
delete(q.members, item.uid)
compactionQueueDepth.Dec()
return &item, true
}
type CompactorState struct {
latestUID models.Uid
latestDID string
status string
stats *carstore.CompactionStats
}
// Compactor is a compactor daemon that compacts repos in the background
type Compactor struct {
q *uniQueue
state *CompactorState
stateLk sync.RWMutex
exit chan struct{}
exited chan struct{}
requeueInterval time.Duration
requeueLimit int
requeueShardCount int
requeueFast bool
}
type CompactorOptions struct {
RequeueInterval time.Duration
RequeueLimit int
RequeueShardCount int
RequeueFast bool
}
func DefaultCompactorOptions() *CompactorOptions {
return &CompactorOptions{
RequeueInterval: time.Hour * 4,
RequeueLimit: 0,
RequeueShardCount: 50,
RequeueFast: true,
}
}
func NewCompactor(opts *CompactorOptions) *Compactor {
if opts == nil {
opts = DefaultCompactorOptions()
}
return &Compactor{
q: &uniQueue{
members: make(map[models.Uid]struct{}),
},
state: &CompactorState{},
exit: make(chan struct{}),
exited: make(chan struct{}),
requeueInterval: opts.RequeueInterval,
requeueLimit: opts.RequeueLimit,
requeueFast: opts.RequeueFast,
requeueShardCount: opts.RequeueShardCount,
}
}
type compactionStats struct {
Completed map[models.Uid]*carstore.CompactionStats
Targets []carstore.CompactionTarget
}
func (c *Compactor) SetState(uid models.Uid, did, status string, stats *carstore.CompactionStats) {
c.stateLk.Lock()
defer c.stateLk.Unlock()
c.state.latestUID = uid
c.state.latestDID = did
c.state.status = status
c.state.stats = stats
}
func (c *Compactor) GetState() *CompactorState {
c.stateLk.RLock()
defer c.stateLk.RUnlock()
return &CompactorState{
latestUID: c.state.latestUID,
latestDID: c.state.latestDID,
status: c.state.status,
stats: c.state.stats,
}
}
var errNoReposToCompact = fmt.Errorf("no repos to compact")
// Start starts the compactor
func (c *Compactor) Start(bgs *BGS) {
log.Info("starting compactor")
go c.doWork(bgs)
if c.requeueInterval > 0 {
go func() {
log.Infow("starting compactor requeue routine",
"interval", c.requeueInterval,
"limit", c.requeueLimit,
"shardCount", c.requeueShardCount,
"fast", c.requeueFast,
)
t := time.NewTicker(c.requeueInterval)
for {
select {
case <-c.exit:
return
case <-t.C:
ctx := context.Background()
ctx, span := otel.Tracer("compactor").Start(ctx, "RequeueRoutine")
if err := c.EnqueueAllRepos(ctx, bgs, c.requeueLimit, c.requeueShardCount, c.requeueFast); err != nil {
log.Errorw("failed to enqueue all repos", "err", err)
}
span.End()
}
}
}()
}
}
// Shutdown shuts down the compactor
func (c *Compactor) Shutdown() {
log.Info("stopping compactor")
close(c.exit)
<-c.exited
log.Info("compactor stopped")
}
func (c *Compactor) doWork(bgs *BGS) {
for {
select {
case <-c.exit:
log.Info("compactor worker exiting, no more active compactions running")
close(c.exited)
return
default:
}
ctx := context.Background()
start := time.Now()
state, err := c.compactNext(ctx, bgs)
if err != nil {
if err == errNoReposToCompact {
log.Debug("no repos to compact, waiting and retrying")
time.Sleep(time.Second * 5)
continue
}
state = c.GetState()
log.Errorw("failed to compact repo",
"err", err,
"uid", state.latestUID,
"repo", state.latestDID,
"status", state.status,
"stats", state.stats,
"duration", time.Since(start),
)
// Pause for a bit to avoid spamming failed compactions
time.Sleep(time.Millisecond * 100)
} else {
log.Infow("compacted repo",
"uid", state.latestUID,
"repo", state.latestDID,
"status", state.status,
"stats", state.stats,
"duration", time.Since(start),
)
}
}
}
func (c *Compactor) compactNext(ctx context.Context, bgs *BGS) (*CompactorState, error) {
ctx, span := otel.Tracer("compactor").Start(ctx, "CompactNext")
defer span.End()
item, ok := c.q.Pop()
if !ok || item == nil {
return nil, errNoReposToCompact
}
c.SetState(item.uid, "unknown", "getting_user", nil)
user, err := bgs.lookupUserByUID(ctx, item.uid)
if err != nil {
span.RecordError(err)
c.SetState(item.uid, "unknown", "failed_getting_user", nil)
return nil, fmt.Errorf("failed to get user %d: %w", item.uid, err)
}
span.SetAttributes(attribute.String("repo", user.Did), attribute.Int("uid", int(item.uid)))
c.SetState(item.uid, user.Did, "compacting", nil)
start := time.Now()
st, err := bgs.repoman.CarStore().CompactUserShards(ctx, item.uid, item.fast)
if err != nil {
span.RecordError(err)
c.SetState(item.uid, user.Did, "failed_compacting", nil)
return nil, fmt.Errorf("failed to compact shards for user %d: %w", item.uid, err)
}
compactionDuration.Observe(time.Since(start).Seconds())
span.SetAttributes(
attribute.Int("shards.deleted", st.ShardsDeleted),
attribute.Int("shards.new", st.NewShards),
attribute.Int("dupes", st.DupeCount),
attribute.Int("shards.skipped", st.SkippedShards),
attribute.Int("refs", st.TotalRefs),
)
c.SetState(item.uid, user.Did, "done", st)
return c.GetState(), nil
}
func (c *Compactor) EnqueueRepo(ctx context.Context, user User, fast bool) {
ctx, span := otel.Tracer("compactor").Start(ctx, "EnqueueRepo")
defer span.End()
log.Infow("enqueueing compaction for repo", "repo", user.Did, "uid", user.ID, "fast", fast)
c.q.Append(user.ID, fast)
}
// EnqueueAllRepos enqueues all repos for compaction
// lim is the maximum number of repos to enqueue
// shardCount is the number of shards to compact per user (0 = default of 50)
// fast is whether to use the fast compaction method (skip large shards)
func (c *Compactor) EnqueueAllRepos(ctx context.Context, bgs *BGS, lim int, shardCount int, fast bool) error {
ctx, span := otel.Tracer("compactor").Start(ctx, "EnqueueAllRepos")
defer span.End()
span.SetAttributes(
attribute.Int("lim", lim),
attribute.Int("shardCount", shardCount),
attribute.Bool("fast", fast),
)
if shardCount == 0 {
shardCount = 20
}
span.SetAttributes(attribute.Int("clampedShardCount", shardCount))
log := log.With("source", "compactor_enqueue_all_repos", "lim", lim, "shardCount", shardCount, "fast", fast)
log.Info("enqueueing all repos")
repos, err := bgs.repoman.CarStore().GetCompactionTargets(ctx, shardCount)
if err != nil {
return fmt.Errorf("failed to get repos to compact: %w", err)
}
span.SetAttributes(attribute.Int("repos", len(repos)))
if lim > 0 && len(repos) > lim {
repos = repos[:lim]
}
span.SetAttributes(attribute.Int("clampedRepos", len(repos)))
for _, r := range repos {
c.q.Append(r.Usr, fast)
}
log.Infow("done enqueueing all repos", "repos_enqueued", len(repos))
return nil
}