-
Notifications
You must be signed in to change notification settings - Fork 282
/
stateless.go
566 lines (492 loc) · 14.5 KB
/
stateless.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
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
// Package stateless implements a PinTracker component for IPFS Cluster, which
// aims to reduce the memory footprint when handling really large cluster
// states.
package stateless
import (
"context"
"errors"
"sync"
"time"
"github.com/ipfs/ipfs-cluster/api"
"github.com/ipfs/ipfs-cluster/pintracker/optracker"
"github.com/ipfs/ipfs-cluster/state"
cid "github.com/ipfs/go-cid"
logging "github.com/ipfs/go-log/v2"
peer "github.com/libp2p/go-libp2p-core/peer"
rpc "github.com/libp2p/go-libp2p-gorpc"
"go.opencensus.io/trace"
)
var logger = logging.Logger("pintracker")
var (
// ErrFullQueue is the error used when pin or unpin operation channel is full.
ErrFullQueue = errors.New("pin/unpin operation queue is full. Try increasing max_pin_queue_size")
// items with this error should be recovered
errUnexpectedlyUnpinned = errors.New("the item should be pinned but it is not")
)
// Tracker uses the optracker.OperationTracker to manage
// transitioning shared ipfs-cluster state (Pins) to the local IPFS node.
type Tracker struct {
config *Config
optracker *optracker.OperationTracker
peerID peer.ID
peerName string
ctx context.Context
cancel func()
getState func(ctx context.Context) (state.ReadOnly, error)
rpcClient *rpc.Client
rpcReady chan struct{}
pinCh chan *optracker.Operation
unpinCh chan *optracker.Operation
shutdownMu sync.Mutex
shutdown bool
wg sync.WaitGroup
}
// New creates a new StatelessPinTracker.
func New(cfg *Config, pid peer.ID, peerName string, getState func(ctx context.Context) (state.ReadOnly, error)) *Tracker {
ctx, cancel := context.WithCancel(context.Background())
spt := &Tracker{
config: cfg,
peerID: pid,
peerName: peerName,
ctx: ctx,
cancel: cancel,
getState: getState,
optracker: optracker.NewOperationTracker(ctx, pid, peerName),
rpcReady: make(chan struct{}, 1),
pinCh: make(chan *optracker.Operation, cfg.MaxPinQueueSize),
unpinCh: make(chan *optracker.Operation, cfg.MaxPinQueueSize),
}
for i := 0; i < spt.config.ConcurrentPins; i++ {
go spt.opWorker(spt.pin, spt.pinCh)
}
go spt.opWorker(spt.unpin, spt.unpinCh)
return spt
}
// receives a pin Function (pin or unpin) and a channel.
// Used for both pinning and unpinning
func (spt *Tracker) opWorker(pinF func(*optracker.Operation) error, opChan chan *optracker.Operation) {
for {
select {
case op := <-opChan:
if cont := applyPinF(pinF, op); cont {
continue
}
spt.optracker.Clean(op.Context(), op)
case <-spt.ctx.Done():
return
}
}
}
// applyPinF returns true if caller should call `continue` inside calling loop.
func applyPinF(pinF func(*optracker.Operation) error, op *optracker.Operation) bool {
if op.Cancelled() {
// operation was cancelled. Move on.
// This saves some time, but not 100% needed.
return true
}
op.SetPhase(optracker.PhaseInProgress)
err := pinF(op) // call pin/unpin
if err != nil {
if op.Cancelled() {
// there was an error because
// we were cancelled. Move on.
return true
}
op.SetError(err)
op.Cancel()
return true
}
op.SetPhase(optracker.PhaseDone)
op.Cancel()
return false
}
func (spt *Tracker) pin(op *optracker.Operation) error {
ctx, span := trace.StartSpan(op.Context(), "tracker/stateless/pin")
defer span.End()
logger.Debugf("issuing pin call for %s", op.Cid())
err := spt.rpcClient.CallContext(
ctx,
"",
"IPFSConnector",
"Pin",
op.Pin(),
&struct{}{},
)
if err != nil {
return err
}
return nil
}
func (spt *Tracker) unpin(op *optracker.Operation) error {
ctx, span := trace.StartSpan(op.Context(), "tracker/stateless/unpin")
defer span.End()
logger.Debugf("issuing unpin call for %s", op.Cid())
err := spt.rpcClient.CallContext(
ctx,
"",
"IPFSConnector",
"Unpin",
op.Pin(),
&struct{}{},
)
if err != nil {
return err
}
return nil
}
// Enqueue puts a new operation on the queue, unless ongoing exists.
func (spt *Tracker) enqueue(ctx context.Context, c *api.Pin, typ optracker.OperationType) error {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/enqueue")
defer span.End()
logger.Debugf("entering enqueue: pin: %+v", c)
op := spt.optracker.TrackNewOperation(ctx, c, typ, optracker.PhaseQueued)
if op == nil {
return nil // ongoing operation.
}
var ch chan *optracker.Operation
switch typ {
case optracker.OperationPin:
ch = spt.pinCh
case optracker.OperationUnpin:
ch = spt.unpinCh
}
select {
case ch <- op:
default:
err := ErrFullQueue
op.SetError(err)
op.Cancel()
logger.Error(err.Error())
return err
}
return nil
}
// SetClient makes the StatelessPinTracker ready to perform RPC requests to
// other components.
func (spt *Tracker) SetClient(c *rpc.Client) {
spt.rpcClient = c
spt.rpcReady <- struct{}{}
}
// Shutdown finishes the services provided by the StatelessPinTracker
// and cancels any active context.
func (spt *Tracker) Shutdown(ctx context.Context) error {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/Shutdown")
_ = ctx
defer span.End()
spt.shutdownMu.Lock()
defer spt.shutdownMu.Unlock()
if spt.shutdown {
logger.Debug("already shutdown")
return nil
}
logger.Info("stopping StatelessPinTracker")
spt.cancel()
close(spt.rpcReady)
spt.wg.Wait()
spt.shutdown = true
return nil
}
// Track tells the StatelessPinTracker to start managing a Cid,
// possibly triggering Pin operations on the IPFS daemon.
func (spt *Tracker) Track(ctx context.Context, c *api.Pin) error {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/Track")
defer span.End()
logger.Debugf("tracking %s", c.Cid)
// Sharded pins are never pinned. A sharded pin cannot turn into
// something else or viceversa like it happens with Remote pins so
// we just ignore them.
if c.Type == api.MetaType {
return nil
}
// Trigger unpin whenever something remote is tracked
// Note, IPFSConn checks with pin/ls before triggering
// pin/rm.
if c.IsRemotePin(spt.peerID) {
op := spt.optracker.TrackNewOperation(ctx, c, optracker.OperationRemote, optracker.PhaseInProgress)
if op == nil {
return nil // ongoing unpin
}
err := spt.unpin(op)
op.Cancel()
if err != nil {
op.SetError(err)
return nil
}
op.SetPhase(optracker.PhaseDone)
spt.optracker.Clean(ctx, op)
return nil
}
return spt.enqueue(ctx, c, optracker.OperationPin)
}
// Untrack tells the StatelessPinTracker to stop managing a Cid.
// If the Cid is pinned locally, it will be unpinned.
func (spt *Tracker) Untrack(ctx context.Context, c cid.Cid) error {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/Untrack")
defer span.End()
logger.Debugf("untracking %s", c)
return spt.enqueue(ctx, api.PinCid(c), optracker.OperationUnpin)
}
// StatusAll returns information for all Cids pinned to the local IPFS node.
func (spt *Tracker) StatusAll(ctx context.Context, filter api.TrackerStatus) []*api.PinInfo {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/StatusAll")
defer span.End()
pininfos, err := spt.localStatus(ctx, true, filter)
if err != nil {
return nil
}
// get all inflight operations from optracker and put them into the
// map, deduplicating any existing items with their inflight operation.
//
// we cannot filter in GetAll, because we are meant to replace items in
// pininfos and set the correct status, as otherwise they will remain in
// PinError.
for _, infop := range spt.optracker.GetAll(ctx) {
pininfos[infop.Cid] = infop
}
var pis []*api.PinInfo
for _, pi := range pininfos {
// Last filter.
if pi.Status.Match(filter) {
pis = append(pis, pi)
}
}
return pis
}
// Status returns information for a Cid pinned to the local IPFS node.
func (spt *Tracker) Status(ctx context.Context, c cid.Cid) *api.PinInfo {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/Status")
defer span.End()
// check if c has an inflight operation or errorred operation in optracker
if oppi, ok := spt.optracker.GetExists(ctx, c); ok {
// if it does return the status of the operation
return oppi
}
pinInfo := &api.PinInfo{
Cid: c,
Peer: spt.peerID,
PinInfoShort: api.PinInfoShort{
PeerName: spt.peerName,
TS: time.Now(),
},
}
// check global state to see if cluster should even be caring about
// the provided cid
var gpin *api.Pin
st, err := spt.getState(ctx)
if err != nil {
logger.Error(err)
addError(pinInfo, err)
return pinInfo
}
gpin, err = st.Get(ctx, c)
if err == state.ErrNotFound {
pinInfo.Status = api.TrackerStatusUnpinned
return pinInfo
}
if err != nil {
logger.Error(err)
addError(pinInfo, err)
return pinInfo
}
// The pin IS in the state.
pinInfo.Name = gpin.Name
// check if pin is a meta pin
if gpin.Type == api.MetaType {
pinInfo.Status = api.TrackerStatusSharded
return pinInfo
}
// check if pin is a remote pin
if gpin.IsRemotePin(spt.peerID) {
pinInfo.Status = api.TrackerStatusRemote
return pinInfo
}
// else attempt to get status from ipfs node
var ips api.IPFSPinStatus
err = spt.rpcClient.CallContext(
ctx,
"",
"IPFSConnector",
"PinLsCid",
gpin,
&ips,
)
if err != nil {
logger.Error(err)
addError(pinInfo, err)
return pinInfo
}
ipfsStatus := ips.ToTrackerStatus()
switch ipfsStatus {
case api.TrackerStatusUnpinned:
// The item is in the state but not in IPFS:
// PinError. Should be pinned.
pinInfo.Status = api.TrackerStatusPinError
pinInfo.Error = errUnexpectedlyUnpinned.Error()
default:
pinInfo.Status = ipfsStatus
}
return pinInfo
}
// RecoverAll attempts to recover all items tracked by this peer.
func (spt *Tracker) RecoverAll(ctx context.Context) ([]*api.PinInfo, error) {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/RecoverAll")
defer span.End()
statuses := spt.StatusAll(ctx, api.TrackerStatusUndefined)
resp := make([]*api.PinInfo, 0)
for _, st := range statuses {
r, err := spt.recoverWithPinInfo(ctx, st)
if err != nil {
return resp, err
}
resp = append(resp, r)
}
return resp, nil
}
// Recover will trigger pinning or unpinning for items in
// PinError or UnpinError states.
func (spt *Tracker) Recover(ctx context.Context, c cid.Cid) (*api.PinInfo, error) {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/Recover")
defer span.End()
// Check if we have a status in the operation tracker
pi, ok := spt.optracker.GetExists(ctx, c)
if ok {
return spt.recoverWithPinInfo(ctx, pi)
}
// Get a status by checking against IPFS and use that.
return spt.recoverWithPinInfo(ctx, spt.Status(ctx, c))
}
func (spt *Tracker) recoverWithPinInfo(ctx context.Context, pi *api.PinInfo) (*api.PinInfo, error) {
var err error
switch pi.Status {
case api.TrackerStatusPinError, api.TrackerStatusUnexpectedlyUnpinned:
logger.Infof("Restarting pin operation for %s", pi.Cid)
err = spt.enqueue(ctx, api.PinCid(pi.Cid), optracker.OperationPin)
case api.TrackerStatusUnpinError:
logger.Infof("Restarting unpin operation for %s", pi.Cid)
err = spt.enqueue(ctx, api.PinCid(pi.Cid), optracker.OperationUnpin)
}
if err != nil {
return spt.Status(ctx, pi.Cid), err
}
return spt.Status(ctx, pi.Cid), nil
}
func (spt *Tracker) ipfsStatusAll(ctx context.Context) (map[cid.Cid]*api.PinInfo, error) {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/ipfsStatusAll")
defer span.End()
var ipsMap map[string]api.IPFSPinStatus
err := spt.rpcClient.CallContext(
ctx,
"",
"IPFSConnector",
"PinLs",
"recursive",
&ipsMap,
)
if err != nil {
logger.Error(err)
return nil, err
}
pins := make(map[cid.Cid]*api.PinInfo, len(ipsMap))
for cidstr, ips := range ipsMap {
c, err := cid.Decode(cidstr)
if err != nil {
logger.Error(err)
continue
}
p := &api.PinInfo{
Cid: c,
Name: "", // to be filled later
Peer: spt.peerID,
PinInfoShort: api.PinInfoShort{
PeerName: spt.peerName,
Status: ips.ToTrackerStatus(),
TS: time.Now(),
},
}
pins[c] = p
}
return pins, nil
}
// localStatus returns a joint set of consensusState and ipfsStatus marking
// pins which should be meta or remote and leaving any ipfs pins that aren't
// in the consensusState out. If incExtra is true, Remote and Sharded pins
// will be added to the status slice. If a filter is provided, only statuses
// matching the filter will be returned.
func (spt *Tracker) localStatus(ctx context.Context, incExtra bool, filter api.TrackerStatus) (map[cid.Cid]*api.PinInfo, error) {
ctx, span := trace.StartSpan(ctx, "tracker/stateless/localStatus")
defer span.End()
// get shared state
st, err := spt.getState(ctx)
if err != nil {
logger.Error(err)
return nil, err
}
statePins, err := st.List(ctx)
if err != nil {
logger.Error(err)
return nil, err
}
var localpis map[cid.Cid]*api.PinInfo
// Only query IPFS if we want to status for pinned items
if filter.Match(api.TrackerStatusPinned | api.TrackerStatusUnexpectedlyUnpinned) {
localpis, err = spt.ipfsStatusAll(ctx)
if err != nil {
logger.Error(err)
return nil, err
}
}
pininfos := make(map[cid.Cid]*api.PinInfo, len(statePins))
for _, p := range statePins {
ipfsInfo, pinnedInIpfs := localpis[p.Cid]
// base pinInfo object - status to be filled.
pinInfo := api.PinInfo{
Cid: p.Cid,
Name: p.Name,
Peer: spt.peerID,
PinInfoShort: api.PinInfoShort{
PeerName: spt.peerName,
TS: time.Now(),
},
}
switch {
case p.Type == api.MetaType:
if !incExtra || !filter.Match(api.TrackerStatusSharded) {
continue
}
pinInfo.Status = api.TrackerStatusSharded
pininfos[p.Cid] = &pinInfo
case p.IsRemotePin(spt.peerID):
if !incExtra || !filter.Match(api.TrackerStatusRemote) {
continue
}
pinInfo.Status = api.TrackerStatusRemote
pininfos[p.Cid] = &pinInfo
case pinnedInIpfs: // always false unless filter matches TrackerStatusPinnned
ipfsInfo.Name = p.Name
pininfos[p.Cid] = ipfsInfo
default:
// report as UNEXPECTEDLY_UNPINNED for this peer.
// this will be overwritten if the operation tracker
// has more info for this (an ongoing pinning
// operation). Otherwise, it means something should be
// pinned and it is not known by IPFS. Should be
// handled to "recover".
pinInfo.Status = api.TrackerStatusUnexpectedlyUnpinned
pinInfo.Error = errUnexpectedlyUnpinned.Error()
pininfos[p.Cid] = &pinInfo
}
}
return pininfos, nil
}
// func (spt *Tracker) getErrorsAll(ctx context.Context) []*api.PinInfo {
// return spt.optracker.Filter(ctx, optracker.PhaseError)
// }
// OpContext exports the internal optracker's OpContext method.
// For testing purposes only.
func (spt *Tracker) OpContext(ctx context.Context, c cid.Cid) context.Context {
return spt.optracker.OpContext(ctx, c)
}
func addError(pinInfo *api.PinInfo, err error) {
pinInfo.Error = err.Error()
pinInfo.Status = api.TrackerStatusClusterError
}