1
0
mirror of https://github.com/distribution/distribution synced 2024-11-12 05:45:51 +01:00
distribution/notifications/sinks.go

165 lines
3.7 KiB
Go
Raw Normal View History

Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
package notifications
import (
"container/list"
"fmt"
"sync"
events "github.com/docker/go-events"
"github.com/sirupsen/logrus"
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
)
// eventQueue accepts all messages into a queue for asynchronous consumption
// by a sink. It is unbounded and thread safe but the sink must be reliable or
// events will be dropped.
type eventQueue struct {
sink events.Sink
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
events *list.List
listeners []eventQueueListener
cond *sync.Cond
mu sync.Mutex
closed bool
}
// eventQueueListener is called when various events happen on the queue.
type eventQueueListener interface {
ingress(event events.Event)
egress(event events.Event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
}
// newEventQueue returns a queue to the provided sink. If the updater is non-
// nil, it will be called to update pending metrics on ingress and egress.
func newEventQueue(sink events.Sink, listeners ...eventQueueListener) *eventQueue {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
eq := eventQueue{
sink: sink,
events: list.New(),
listeners: listeners,
}
eq.cond = sync.NewCond(&eq.mu)
go eq.run()
return &eq
}
// Write accepts the events into the queue, only failing if the queue has
// beend closed.
func (eq *eventQueue) Write(event events.Event) error {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
eq.mu.Lock()
defer eq.mu.Unlock()
if eq.closed {
return ErrSinkClosed
}
for _, listener := range eq.listeners {
listener.ingress(event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
}
eq.events.PushBack(event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
eq.cond.Signal() // signal waiters
return nil
}
// Close shuts down the event queue, flushing
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
func (eq *eventQueue) Close() error {
eq.mu.Lock()
defer eq.mu.Unlock()
if eq.closed {
return fmt.Errorf("eventqueue: already closed")
}
// set closed flag
eq.closed = true
eq.cond.Signal() // signal flushes queue
eq.cond.Wait() // wait for signal from last flush
return eq.sink.Close()
}
// run is the main goroutine to flush events to the target sink.
func (eq *eventQueue) run() {
for {
event := eq.next()
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
if event == nil {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
return // nil block means event queue is closed.
}
if err := eq.sink.Write(event); err != nil {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
logrus.Warnf("eventqueue: error writing events to %v, these events will be lost: %v", eq.sink, err)
}
for _, listener := range eq.listeners {
listener.egress(event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
}
}
}
// next encompasses the critical section of the run loop. When the queue is
// empty, it will block on the condition. If new data arrives, it will wake
// and return a block. When closed, a nil slice will be returned.
func (eq *eventQueue) next() events.Event {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
eq.mu.Lock()
defer eq.mu.Unlock()
for eq.events.Len() < 1 {
if eq.closed {
eq.cond.Broadcast()
return nil
}
eq.cond.Wait()
}
front := eq.events.Front()
block := front.Value.(events.Event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
eq.events.Remove(front)
return block
}
// ignoredSink discards events with ignored target media types and actions.
// passes the rest along.
type ignoredSink struct {
events.Sink
ignoreMediaTypes map[string]bool
ignoreActions map[string]bool
}
func newIgnoredSink(sink events.Sink, ignored []string, ignoreActions []string) events.Sink {
if len(ignored) == 0 {
return sink
}
ignoredMap := make(map[string]bool)
for _, mediaType := range ignored {
ignoredMap[mediaType] = true
}
ignoredActionsMap := make(map[string]bool)
for _, action := range ignoreActions {
ignoredActionsMap[action] = true
}
return &ignoredSink{
Sink: sink,
ignoreMediaTypes: ignoredMap,
ignoreActions: ignoredActionsMap,
}
}
// Write discards events with ignored target media types and passes the rest
// along.
func (imts *ignoredSink) Write(event events.Event) error {
if imts.ignoreMediaTypes[event.(Event).Target.MediaType] || imts.ignoreActions[event.(Event).Action] {
return nil
}
return imts.Sink.Write(event)
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
}
func (imts *ignoredSink) Close() error {
Implement notification endpoint webhook dispatch This changeset implements webhook notification endpoints for dispatching registry events. Repository instances can be decorated by a listener that converts calls into context-aware events, using a bridge. Events generated in the bridge are written to a sink. Implementations of sink include a broadcast and endpoint sink which can be used to configure event dispatch. Endpoints represent a webhook notification target, with queueing and retries built in. They can be added to a Broadcaster, which is a simple sink that writes a block of events to several sinks, to provide a complete dispatch mechanism. The main caveat to the current approach is that all unsent notifications are inmemory. Best effort is made to ensure that notifications are not dropped, to the point where queues may back up on faulty endpoints. If the endpoint is fixed, the events will be retried and all messages will go through. Internally, this functionality is all made up of Sink objects. The queuing functionality is implemented with an eventQueue sink and retries are implemented with retryingSink. Replacing the inmemory queuing with something persistent should be as simple as replacing broadcaster with a remote queue and that sets up the sinks to be local workers listening to that remote queue. Metrics are kept for each endpoint and exported via expvar. This may not be a permanent appraoch but should provide enough information for troubleshooting notification problems. Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-28 08:27:46 +01:00
return nil
}