summaryrefslogtreecommitdiff
path: root/libpod/pod.go
blob: 9ea1f058bd4d75e9738b1c53b899ed03fa07ce4f (plain)
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
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
package libpod

import (
	"context"
	"path/filepath"
	"strings"
	"time"

	"github.com/containers/storage"
	"github.com/docker/docker/pkg/stringid"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
)

// Pod represents a group of containers that are managed together.
// Any operations on a Pod that access state must begin with a call to
// updatePod().
// There is no guarantee that state exists in a readable state before this call,
// and even if it does its contents will be out of date and must be refreshed
// from the database.
// Generally, this requirement applies only to top-level functions; helpers can
// assume their callers handled this requirement. Generally speaking, if a
// function takes the pod lock and accesses any part of state, it should
// updatePod() immediately after locking.
// ffjson: skip
type Pod struct {
	config *PodConfig
	state  *podState

	valid   bool
	runtime *Runtime
	lock    storage.Locker
}

// PodConfig represents a pod's static configuration
type PodConfig struct {
	ID   string `json:"id"`
	Name string `json:"name"`
	// Namespace the pod is in
	Namespace string `json:"namespace,omitempty"`

	// Labels contains labels applied to the pod
	Labels map[string]string `json:"labels"`
	// CgroupParent contains the pod's CGroup parent
	CgroupParent string `json:"cgroupParent"`
	// UsePodCgroup indicates whether the pod will create its own CGroup and
	// join containers to it.
	// If true, all containers joined to the pod will use the pod cgroup as
	// their cgroup parent, and cannot set a different cgroup parent
	UsePodCgroup bool

	// Time pod was created
	CreatedTime time.Time `json:"created"`
}

// podState represents a pod's state
type podState struct {
	// CgroupPath is the path to the pod's CGroup
	CgroupPath string
}

// ID retrieves the pod's ID
func (p *Pod) ID() string {
	return p.config.ID
}

// Name retrieves the pod's name
func (p *Pod) Name() string {
	return p.config.Name
}

// Namespace returns the pod's libpod namespace.
// Namespaces are used to logically separate containers and pods in the state.
func (p *Pod) Namespace() string {
	return p.config.Namespace
}

// Labels returns the pod's labels
func (p *Pod) Labels() map[string]string {
	labels := make(map[string]string)
	for key, value := range p.config.Labels {
		labels[key] = value
	}

	return labels
}

// CreatedTime gets the time when the pod was created
func (p *Pod) CreatedTime() time.Time {
	return p.config.CreatedTime
}

// CgroupParent returns the pod's CGroup parent
func (p *Pod) CgroupParent() string {
	return p.config.CgroupParent
}

// UsePodCgroup returns whether containers in the pod will default to this pod's
// cgroup instead of the default libpod parent
func (p *Pod) UsePodCgroup() bool {
	return p.config.UsePodCgroup
}

// CgroupPath returns the path to the pod's CGroup
func (p *Pod) CgroupPath() (string, error) {
	p.lock.Lock()
	defer p.lock.Unlock()
	if err := p.updatePod(); err != nil {
		return "", err
	}

	return p.state.CgroupPath, nil
}

// Creates a new, empty pod
func newPod(lockDir string, runtime *Runtime) (*Pod, error) {
	pod := new(Pod)
	pod.config = new(PodConfig)
	pod.config.ID = stringid.GenerateNonCryptoID()
	pod.config.Labels = make(map[string]string)
	pod.config.CreatedTime = time.Now()
	pod.state = new(podState)
	pod.runtime = runtime

	// Path our lock file will reside at
	lockPath := filepath.Join(lockDir, pod.config.ID)
	// Grab a lockfile at the given path
	lock, err := storage.GetLockfile(lockPath)
	if err != nil {
		return nil, errors.Wrapf(err, "error creating lockfile for new pod")
	}
	pod.lock = lock

	return pod, nil
}

// Update pod state from database
func (p *Pod) updatePod() error {
	if err := p.runtime.state.UpdatePod(p); err != nil {
		return err
	}

	return nil
}

// Save pod state to database
func (p *Pod) save() error {
	if err := p.runtime.state.SavePod(p); err != nil {
		return errors.Wrapf(err, "error saving pod %s state")
	}

	return nil
}

// Refresh a pod's state after restart
func (p *Pod) refresh() error {
	if !p.valid {
		return ErrPodRemoved
	}

	// We need to recreate the pod's cgroup
	if p.config.UsePodCgroup {
		switch p.runtime.config.CgroupManager {
		case SystemdCgroupsManager:
			// NOOP for now, until proper systemd cgroup management
			// is implemented
		case CgroupfsCgroupsManager:
			p.state.CgroupPath = filepath.Join(p.config.CgroupParent, p.ID())

			logrus.Debugf("setting pod cgroup to %s", p.state.CgroupPath)
		default:
			return errors.Wrapf(ErrInvalidArg, "unknown cgroups manager %s specified", p.runtime.config.CgroupManager)
		}
	}

	// Save changes
	return p.save()
}

// Start starts all containers within a pod
// It combines the effects of Init() and Start() on a container
// If a container has already been initialized it will be started,
// otherwise it will be initialized then started.
// Containers that are already running or have been paused are ignored
// All containers are started independently, in order dictated by their
// dependencies.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were started
// If map is not nil, an error was encountered when starting one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were started successfully
func (p *Pod) Start(ctx context.Context) (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	// Build a dependency graph of containers in the pod
	graph, err := buildContainerGraph(allCtrs)
	if err != nil {
		return nil, errors.Wrapf(err, "error generating dependency graph for pod %s", p.ID())
	}

	ctrErrors := make(map[string]error)
	ctrsVisited := make(map[string]bool)

	// If there are no containers without dependencies, we can't start
	// Error out
	if len(graph.noDepNodes) == 0 {
		return nil, errors.Wrapf(ErrNoSuchCtr, "no containers in pod %s have no dependencies, cannot start pod", p.ID())
	}

	// Traverse the graph beginning at nodes with no dependencies
	for _, node := range graph.noDepNodes {
		startNode(ctx, node, false, ctrErrors, ctrsVisited, false)
	}

	return ctrErrors, nil
}

// Visit a node on a container graph and start the container, or set an error if
// a dependency failed to start. if restart is true, startNode will restart the node instead of starting it.
func startNode(ctx context.Context, node *containerNode, setError bool, ctrErrors map[string]error, ctrsVisited map[string]bool, restart bool) {
	// First, check if we have already visited the node
	if ctrsVisited[node.id] {
		return
	}

	// If setError is true, a dependency of us failed
	// Mark us as failed and recurse
	if setError {
		// Mark us as visited, and set an error
		ctrsVisited[node.id] = true
		ctrErrors[node.id] = errors.Wrapf(ErrCtrStateInvalid, "a dependency of container %s failed to start", node.id)

		// Hit anyone who depends on us, and set errors on them too
		for _, successor := range node.dependedOn {
			startNode(ctx, successor, true, ctrErrors, ctrsVisited, restart)
		}

		return
	}

	// Have all our dependencies started?
	// If not, don't visit the node yet
	depsVisited := true
	for _, dep := range node.dependsOn {
		depsVisited = depsVisited && ctrsVisited[dep.id]
	}
	if !depsVisited {
		// Don't visit us yet, all dependencies are not up
		// We'll hit the dependencies eventually, and when we do it will
		// recurse here
		return
	}

	// Going to try to start the container, mark us as visited
	ctrsVisited[node.id] = true

	ctrErrored := false

	// Check if dependencies are running
	// Graph traversal means we should have started them
	// But they could have died before we got here
	// Does not require that the container be locked, we only need to lock
	// the dependencies
	depsStopped, err := node.container.checkDependenciesRunning()
	if err != nil {
		ctrErrors[node.id] = err
		ctrErrored = true
	} else if len(depsStopped) > 0 {
		// Our dependencies are not running
		depsList := strings.Join(depsStopped, ",")
		ctrErrors[node.id] = errors.Wrapf(ErrCtrStateInvalid, "the following dependencies of container %s are not running: %s", node.id, depsList)
		ctrErrored = true
	}

	// Lock before we start
	node.container.lock.Lock()

	// Sync the container to pick up current state
	if !ctrErrored {
		if err := node.container.syncContainer(); err != nil {
			ctrErrored = true
			ctrErrors[node.id] = err
		}
	}

	// Start the container (only if it is not running)
	if !ctrErrored {
		if !restart && node.container.state.State != ContainerStateRunning {
			if err := node.container.initAndStart(ctx); err != nil {
				ctrErrored = true
				ctrErrors[node.id] = err
			}
		}
		if restart && node.container.state.State != ContainerStatePaused && node.container.state.State != ContainerStateUnknown {
			if err := node.container.restartWithTimeout(ctx, node.container.config.StopTimeout); err != nil {
				ctrErrored = true
				ctrErrors[node.id] = err
			}
		}
	}

	node.container.lock.Unlock()

	// Recurse to anyone who depends on us and start them
	for _, successor := range node.dependedOn {
		startNode(ctx, successor, ctrErrored, ctrErrors, ctrsVisited, restart)
	}

	return
}

// Stop stops all containers within a pod that are not already stopped
// Each container will use its own stop timeout
// Only running containers will be stopped. Paused, stopped, or created
// containers will be ignored.
// If cleanup is true, mounts and network namespaces will be cleaned up after
// the container is stopped.
// All containers are stopped independently. An error stopping one container
// will not prevent other containers being stopped.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were stopped
// If map is not nil, an error was encountered when stopping one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were stopped without error
func (p *Pod) Stop(cleanup bool) (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	ctrErrors := make(map[string]error)

	// TODO: There may be cases where it makes sense to order stops based on
	// dependencies. Should we bother with this?

	// Stop to all containers
	for _, ctr := range allCtrs {
		ctr.lock.Lock()

		if err := ctr.syncContainer(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		// Ignore containers that are not running
		if ctr.state.State != ContainerStateRunning {
			ctr.lock.Unlock()
			continue
		}

		if err := ctr.stop(ctr.config.StopTimeout); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		if cleanup {
			if err := ctr.cleanup(); err != nil {
				ctrErrors[ctr.ID()] = err
			}
		}

		ctr.lock.Unlock()
	}

	if len(ctrErrors) > 0 {
		return ctrErrors, errors.Wrapf(ErrCtrExists, "error stopping some containers")
	}

	return nil, nil
}

// Pause pauses all containers within a pod that are running.
// Only running containers will be paused. Paused, stopped, or created
// containers will be ignored.
// All containers are paused independently. An error pausing one container
// will not prevent other containers being paused.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were paused
// If map is not nil, an error was encountered when pausing one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were paused without error
func (p *Pod) Pause() (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	ctrErrors := make(map[string]error)

	// Pause to all containers
	for _, ctr := range allCtrs {
		ctr.lock.Lock()

		if err := ctr.syncContainer(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		// Ignore containers that are not running
		if ctr.state.State != ContainerStateRunning {
			ctr.lock.Unlock()
			continue
		}

		if err := ctr.pause(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		ctr.lock.Unlock()
	}

	if len(ctrErrors) > 0 {
		return ctrErrors, errors.Wrapf(ErrCtrExists, "error pausing some containers")
	}

	return nil, nil
}

// Unpause unpauses all containers within a pod that are running.
// Only paused containers will be unpaused. Running, stopped, or created
// containers will be ignored.
// All containers are unpaused independently. An error unpausing one container
// will not prevent other containers being unpaused.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were unpaused
// If map is not nil, an error was encountered when unpausing one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were unpaused without error
func (p *Pod) Unpause() (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	ctrErrors := make(map[string]error)

	// Pause to all containers
	for _, ctr := range allCtrs {
		ctr.lock.Lock()

		if err := ctr.syncContainer(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		// Ignore containers that are not paused
		if ctr.state.State != ContainerStatePaused {
			ctr.lock.Unlock()
			continue
		}

		if err := ctr.unpause(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		ctr.lock.Unlock()
	}

	if len(ctrErrors) > 0 {
		return ctrErrors, errors.Wrapf(ErrCtrExists, "error unpausing some containers")
	}

	return nil, nil
}

// Restart restarts all containers within a pod that are not paused or in an error state.
// It combines the effects of Stop() and Start() on a container
// Each container will use its own stop timeout.
// All containers are started independently, in order dictated by their
// dependencies. An error restarting one container
// will not prevent other containers being restarted.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were restarted
// If map is not nil, an error was encountered when restarting one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were restarted without error
func (p *Pod) Restart(ctx context.Context) (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	// Build a dependency graph of containers in the pod
	graph, err := buildContainerGraph(allCtrs)
	if err != nil {
		return nil, errors.Wrapf(err, "error generating dependency graph for pod %s", p.ID())
	}

	ctrErrors := make(map[string]error)
	ctrsVisited := make(map[string]bool)

	// If there are no containers without dependencies, we can't start
	// Error out
	if len(graph.noDepNodes) == 0 {
		return nil, errors.Wrapf(ErrNoSuchCtr, "no containers in pod %s have no dependencies, cannot start pod", p.ID())
	}

	// Traverse the graph beginning at nodes with no dependencies
	for _, node := range graph.noDepNodes {
		startNode(ctx, node, false, ctrErrors, ctrsVisited, true)
	}

	if len(ctrErrors) > 0 {
		return ctrErrors, errors.Wrapf(ErrCtrExists, "error stopping some containers")
	}

	return nil, nil
}

// Kill sends a signal to all running containers within a pod
// Signals will only be sent to running containers. Containers that are not
// running will be ignored. All signals are sent independently, and sending will
// continue even if some containers encounter errors.
// An error and a map[string]error are returned
// If the error is not nil and the map is nil, an error was encountered before
// any containers were signalled
// If map is not nil, an error was encountered when signalling one or more
// containers. The container ID is mapped to the error encountered. The error is
// set to ErrCtrExists
// If both error and the map are nil, all containers were signalled successfully
func (p *Pod) Kill(signal uint) (map[string]error, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	ctrErrors := make(map[string]error)

	// Send a signal to all containers
	for _, ctr := range allCtrs {
		ctr.lock.Lock()

		if err := ctr.syncContainer(); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		// Ignore containers that are not running
		if ctr.state.State != ContainerStateRunning {
			ctr.lock.Unlock()
			continue
		}

		if err := ctr.runtime.ociRuntime.killContainer(ctr, signal); err != nil {
			ctr.lock.Unlock()
			ctrErrors[ctr.ID()] = err
			continue
		}

		logrus.Debugf("Killed container %s with signal %d", ctr.ID(), signal)
	}

	if len(ctrErrors) > 0 {
		return ctrErrors, nil
	}

	return nil, nil
}

// HasContainer checks if a container is present in the pod
func (p *Pod) HasContainer(id string) (bool, error) {
	if !p.valid {
		return false, ErrPodRemoved
	}

	return p.runtime.state.PodHasContainer(p, id)
}

// AllContainersByID returns the container IDs of all the containers in the pod
func (p *Pod) AllContainersByID() ([]string, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	return p.runtime.state.PodContainersByID(p)
}

// AllContainers retrieves the containers in the pod
func (p *Pod) AllContainers() ([]*Container, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	return p.runtime.state.PodContainers(p)
}

// Status gets the status of all containers in the pod
// Returns a map of Container ID to Container Status
func (p *Pod) Status() (map[string]ContainerStatus, error) {
	p.lock.Lock()
	defer p.lock.Unlock()

	if !p.valid {
		return nil, ErrPodRemoved
	}

	allCtrs, err := p.runtime.state.PodContainers(p)
	if err != nil {
		return nil, err
	}

	// We need to lock all the containers
	for _, ctr := range allCtrs {
		ctr.lock.Lock()
		defer ctr.lock.Unlock()
	}

	// Now that all containers are locked, get their status
	status := make(map[string]ContainerStatus, len(allCtrs))
	for _, ctr := range allCtrs {
		if err := ctr.syncContainer(); err != nil {
			return nil, err
		}

		status[ctr.ID()] = ctr.state.State
	}

	return status, nil
}

// TODO add pod batching
// Lock pod to avoid lock contention
// Store and lock all containers (no RemoveContainer in batch guarantees cache will not become stale)