Documentation
¶
Index ¶
Constants ¶
const RestartDelay = 3 * time.Second
RestartDelay holds the length of time that a worker will wait between exiting and restarting.
Variables ¶
var ( ErrTerminateAgent = errors.New("agent should be terminated") ErrRebootMachine = errors.New("machine needs to reboot") ErrShutdownMachine = errors.New("machine needs to shutdown") )
These errors are returned by various specific workers in the hope that they will have some specific effect on the top-level agent running that worker.
It should be clear that they don't belong here, and certainly shouldn't be used as they are today: e.g. a uniter has *no fricking idea* whether its host agent should shut down. A uniter can return ErrUnitDead, and its host might need to respond to that, perhaps by returning an error specific to *its* host; depending on these values punching right through N layers (but only when we want them to!) is kinda terrible.
var ErrDead = errors.New("worker runner is not running")
var ErrKilled = errors.New("worker killed")
ErrKilled can be returned by the PeriodicWorkerCall to signify that the function has returned as a result of a Stop() or Kill() signal and that the function was able to stop cleanly
Functions ¶
func Dead ¶
func Dead(worker Worker) <-chan struct{}
Dead returns a channel that will be closed when the supplied Worker has completed.
Don't be too casual about calling Dead -- for example, in a standard select loop, `case <-worker.Dead(w):` will create one new goroutine per iteration, which is... untidy.
Types ¶
type FinishedWorker ¶
type FinishedWorker struct{}
FinishedWorker is a worker that stops immediately with no error when started by a Runner, which then removes it from the list of workers without restarting it. Simply return FinishedWorker{} where you need to avoid starting a worker at all.
func (FinishedWorker) Kill ¶
func (w FinishedWorker) Kill()
Kill implements Worker.Kill() and does nothing.
func (FinishedWorker) Wait ¶
func (w FinishedWorker) Wait() error
Wait implements Worker.Wait() and immediately returns no error.
type NewTimerFunc ¶
type NewTimerFunc func(time.Duration) PeriodicTimer
NewTimerFunc is a constructor used to obtain the instance of PeriodicTimer periodicWorker uses on its loop. TODO(fwereade): 2016-03-17 lp:1558657
type PeriodicTimer ¶
type PeriodicTimer interface {
// Reset changes the timer to expire after duration d.
// It returns true if the timer had been active, false
// if the timer had expired or been stopped.
Reset(time.Duration) bool
// CountDown returns the channel used to signal expiration of
// the timer duration. The channel is called C in the base
// implementation of timer but the name is confusing.
CountDown() <-chan time.Time
}
PeriodicTimer is an interface for the timer that periodicworker will use to handle the calls.
func NewTimer ¶
func NewTimer(d time.Duration) PeriodicTimer
NewTimer is the default implementation of NewTimerFunc.
type PeriodicWorkerCall ¶
type PeriodicWorkerCall func(stop <-chan struct{}) error
PeriodicWorkerCall represents the callable to be passed to the periodic worker to be run every elapsed period.
type Runner ¶
type Runner interface {
Worker
StartWorker(id string, startFunc func() (Worker, error)) error
StopWorker(id string) error
}
Runner is implemented by instances capable of starting and stopping workers.
func NewRunner ¶
func NewRunner(isFatal func(error) bool, moreImportant func(err0, err1 error) bool, restartDelay time.Duration) Runner
NewRunner creates a new Runner. When a worker finishes, if its error is deemed fatal (determined by calling isFatal), all the other workers will be stopped and the runner itself will finish. Of all the fatal errors returned by the stopped workers, only the most important one, determined by calling moreImportant, will be returned from Runner.Wait. Non-fatal errors will not be returned.
The function isFatal(err) returns whether err is a fatal error. The function moreImportant(err0, err1) returns whether err0 is considered more important than err1.
type Timer ¶
type Timer struct {
// contains filtered or unexported fields
}
Timer implements PeriodicTimer.
type Worker ¶
type Worker interface {
// Kill asks the worker to stop and returns immediately.
Kill()
// Wait waits for the worker to complete and returns any
// error encountered when it was running or stopping.
Wait() error
}
Worker describes any type whose validity and/or activity is bounded in time. Most frequently, they will represent the duration of some task or tasks running on internal goroutines, but it's possible and rational to use them to represent any resource that might become invalid.
Worker implementations must be goroutine-safe.
func NewNoOpWorker ¶
func NewNoOpWorker() Worker
func NewPeriodicWorker ¶
func NewPeriodicWorker(call PeriodicWorkerCall, period time.Duration, timerFunc NewTimerFunc) Worker
NewPeriodicWorker returns a worker that runs the given function continually sleeping for sleepDuration in between each call, until Kill() is called The stopCh argument will be closed when the worker is killed. The error returned by the doWork function will be returned by the worker's Wait function.
func NewSimpleWorker ¶
NewSimpleWorker returns a worker that runs the given function. The stopCh argument will be closed when the worker is killed. The error returned by the doWork function will be returned by the worker's Wait function.
Source Files
¶
Directories
¶
| Path | Synopsis |
|---|---|
|
Catacomb leverages tomb.Tomb to bind the lifetimes of, and track the errors of, a group of related workers.
|
Catacomb leverages tomb.Tomb to bind the lifetimes of, and track the errors of, a group of related workers. |
|
The dependency package exists to address a general problem with shared resources and the management of their lifetimes.
|
The dependency package exists to address a general problem with shared resources and the management of their lifetimes. |
|
Package diskmanager defines a worker that periodically lists block devices on the machine it runs on.
|
Package diskmanager defines a worker that periodically lists block devices on the machine it runs on. |
|
Package fortress implements a convenient metaphor for an RWLock.
|
Package fortress implements a convenient metaphor for an RWLock. |
|
package gate provides a mechanism by which independent workers can wait for one another to finish a task, without introducing explicit dependencies between those workers.
|
package gate provides a mechanism by which independent workers can wait for one another to finish a task, without introducing explicit dependencies between those workers. |
|
Package introspection defines the worker that can report internal agent state through the use of a machine local socket.
|
Package introspection defines the worker that can report internal agent state through the use of a machine local socket. |
|
pprof
Package pprof is a fork of net/http/pprof modified to communicate over a unix socket.
|
Package pprof is a fork of net/http/pprof modified to communicate over a unix socket. |
|
Package meterstatus provides a worker that executes the meter-status-changed hook periodically.
|
Package meterstatus provides a worker that executes the meter-status-changed hook periodically. |
|
metrics
|
|
|
collect
Package collect provides a worker that executes the collect-metrics hook periodically, as long as the workload has been started (between start and stop hooks).
|
Package collect provides a worker that executes the collect-metrics hook periodically, as long as the workload has been started (between start and stop hooks). |
|
sender
Package sender contains the implementation of the metric sender manifold.
|
Package sender contains the implementation of the metric sender manifold. |
|
spool
Package spool contains the implementation of a worker that extracts the spool directory path from the agent config and enables other workers to write and read metrics to and from a the spool directory using a writer and a reader.
|
Package spool contains the implementation of a worker that extracts the spool directory path from the agent config and enables other workers to write and read metrics to and from a the spool directory using a writer and a reader. |
|
Package storageprovisioner provides a worker that manages the provisioning and deprovisioning of storage volumes and filesystems, and attaching them to and detaching them from machines.
|
Package storageprovisioner provides a worker that manages the provisioning and deprovisioning of storage volumes and filesystems, and attaching them to and detaching them from machines. |
|
hook
hook provides types that define the hooks known to the Uniter
|
hook provides types that define the hooks known to the Uniter |
|
relation
relation implements persistent local storage of a unit's relation state, and translation of relation changes into hooks that need to be run.
|
relation implements persistent local storage of a unit's relation state, and translation of relation changes into hooks that need to be run. |
|
runner/context
Package context contains the ContextFactory and Context definitions.
|
Package context contains the ContextFactory and Context definitions. |
|
runner/jujuc
The worker/uniter/runner/jujuc package implements the server side of the jujuc proxy tool, which forwards command invocations to the unit agent process so that they can be executed against specific state.
|
The worker/uniter/runner/jujuc package implements the server side of the jujuc proxy tool, which forwards command invocations to the unit agent process so that they can be executed against specific state. |
|
storage
Package storage contains the storage subsystem for the uniter, responding to changes in storage attachments (lifecycle, volume/filesystem details) by queuing hooks and managing the storage attachments' lifecycle.
|
Package storage contains the storage subsystem for the uniter, responding to changes in storage attachments (lifecycle, volume/filesystem details) by queuing hooks and managing the storage attachments' lifecycle. |