aboutsummaryrefslogtreecommitdiff
path: root/libpod/pod.go
blob: 3cad3f817e1f38d680435bcf9c0d34c44eba00ad (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
package libpod

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

	"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 may share namespaces
// ffjson: skip
type Pod struct {
	config *PodConfig

	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"`
	Labels map[string]string `json:""`
}

// 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
}

// 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
}

// 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.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
}

// 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)
	}

	return ctrErrors, nil
}

// Visit a node on a container graph and start the container, or set an error if
// a dependency failed to start
func startNode(ctx context.Context, node *containerNode, setError bool, ctrErrors map[string]error, ctrsVisited map[string]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)
		}

		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 && node.container.state.State != ContainerStateRunning {
		if err := node.container.initAndStart(ctx); 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)
	}

	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
}

// 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)