mirror of
https://github.com/hibiken/asynq.git
synced 2024-11-10 11:31:58 +08:00
Define broker interface
This commit is contained in:
parent
42453280f4
commit
f8a94fb839
24
asynq.go
24
asynq.go
@ -7,8 +7,10 @@ package asynq
|
||||
import (
|
||||
"crypto/tls"
|
||||
"fmt"
|
||||
"time"
|
||||
|
||||
"github.com/go-redis/redis/v7"
|
||||
"github.com/hibiken/asynq/internal/base"
|
||||
)
|
||||
|
||||
// Task represents a unit of work to be performed.
|
||||
@ -30,6 +32,28 @@ func NewTask(typename string, payload map[string]interface{}) *Task {
|
||||
}
|
||||
}
|
||||
|
||||
// broker is a message broker that supports operations to manage task queues.
|
||||
//
|
||||
// See rdb.RDB as a reference implementation.
|
||||
type broker interface {
|
||||
Enqueue(msg *base.TaskMessage) error
|
||||
EnqueueUnique(msg *base.TaskMessage, ttl time.Duration) error
|
||||
Dequeue(qnames ...string) (*base.TaskMessage, error)
|
||||
Done(msg *base.TaskMessage) error
|
||||
Requeue(msg *base.TaskMessage) error
|
||||
Schedule(msg *base.TaskMessage, processAt time.Time) error
|
||||
ScheduleUnique(msg *base.TaskMessage, processAt time.Time, ttl time.Duration) error
|
||||
Retry(msg *base.TaskMessage, processAt time.Time, errMsg string) error
|
||||
Kill(msg *base.TaskMessage, errMsg string) error
|
||||
RequeueAll() (int64, error)
|
||||
CheckAndEnqueue(qnames ...string) error
|
||||
WriteServerState(ss *base.ServerState, ttl time.Duration) error
|
||||
ClearServerState(ss *base.ServerState) error
|
||||
CancelationPubSub() (*redis.PubSub, error) // TODO: Need to decouple from redis to support other brokers
|
||||
PublishCancelation(id string) error
|
||||
Close() error
|
||||
}
|
||||
|
||||
// RedisConnOpt is a discriminated union of types that represent Redis connection configuration option.
|
||||
//
|
||||
// RedisConnOpt represents a sum of following types:
|
||||
|
11
heartbeat.go
11
heartbeat.go
@ -9,14 +9,13 @@ import (
|
||||
"time"
|
||||
|
||||
"github.com/hibiken/asynq/internal/base"
|
||||
"github.com/hibiken/asynq/internal/rdb"
|
||||
)
|
||||
|
||||
// heartbeater is responsible for writing process info to redis periodically to
|
||||
// indicate that the background worker process is up.
|
||||
type heartbeater struct {
|
||||
logger Logger
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
|
||||
ss *base.ServerState
|
||||
|
||||
@ -27,10 +26,10 @@ type heartbeater struct {
|
||||
interval time.Duration
|
||||
}
|
||||
|
||||
func newHeartbeater(l Logger, rdb *rdb.RDB, ss *base.ServerState, interval time.Duration) *heartbeater {
|
||||
func newHeartbeater(l Logger, b broker, ss *base.ServerState, interval time.Duration) *heartbeater {
|
||||
return &heartbeater{
|
||||
logger: l,
|
||||
rdb: rdb,
|
||||
broker: b,
|
||||
ss: ss,
|
||||
done: make(chan struct{}),
|
||||
interval: interval,
|
||||
@ -53,7 +52,7 @@ func (h *heartbeater) start(wg *sync.WaitGroup) {
|
||||
for {
|
||||
select {
|
||||
case <-h.done:
|
||||
h.rdb.ClearServerState(h.ss)
|
||||
h.broker.ClearServerState(h.ss)
|
||||
h.logger.Info("Heartbeater done")
|
||||
return
|
||||
case <-time.After(h.interval):
|
||||
@ -66,7 +65,7 @@ func (h *heartbeater) start(wg *sync.WaitGroup) {
|
||||
func (h *heartbeater) beat() {
|
||||
// Note: Set TTL to be long enough so that it won't expire before we write again
|
||||
// and short enough to expire quickly once the process is shut down or killed.
|
||||
err := h.rdb.WriteServerState(h.ss, h.interval*2)
|
||||
err := h.broker.WriteServerState(h.ss, h.interval*2)
|
||||
if err != nil {
|
||||
h.logger.Error("could not write heartbeat data: %v", err)
|
||||
}
|
||||
|
26
processor.go
26
processor.go
@ -19,7 +19,7 @@ import (
|
||||
|
||||
type processor struct {
|
||||
logger Logger
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
|
||||
ss *base.ServerState
|
||||
|
||||
@ -65,7 +65,7 @@ type retryDelayFunc func(n int, err error, task *Task) time.Duration
|
||||
|
||||
type newProcessorParams struct {
|
||||
logger Logger
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
ss *base.ServerState
|
||||
retryDelayFunc retryDelayFunc
|
||||
syncCh chan<- *syncRequest
|
||||
@ -84,7 +84,7 @@ func newProcessor(params newProcessorParams) *processor {
|
||||
}
|
||||
return &processor{
|
||||
logger: params.logger,
|
||||
rdb: params.rdb,
|
||||
broker: params.broker,
|
||||
ss: params.ss,
|
||||
queueConfig: qcfg,
|
||||
orderedQueues: orderedQueues,
|
||||
@ -157,8 +157,8 @@ func (p *processor) start(wg *sync.WaitGroup) {
|
||||
// process the task.
|
||||
func (p *processor) exec() {
|
||||
qnames := p.queues()
|
||||
msg, err := p.rdb.Dequeue(qnames...)
|
||||
if err == rdb.ErrNoProcessableTask {
|
||||
msg, err := p.broker.Dequeue(qnames...)
|
||||
if err == rdb.ErrNoProcessableTask { // TODO: Need to decouple this error from rdb to support other brokers
|
||||
// queues are empty, this is a normal behavior.
|
||||
if len(p.queueConfig) > 1 {
|
||||
// sleep to avoid slamming redis and let scheduler move tasks into queues.
|
||||
@ -227,7 +227,7 @@ func (p *processor) exec() {
|
||||
// restore moves all tasks from "in-progress" back to queue
|
||||
// to restore all unfinished tasks.
|
||||
func (p *processor) restore() {
|
||||
n, err := p.rdb.RequeueAll()
|
||||
n, err := p.broker.RequeueAll()
|
||||
if err != nil {
|
||||
p.logger.Error("Could not restore unfinished tasks: %v", err)
|
||||
}
|
||||
@ -237,20 +237,20 @@ func (p *processor) restore() {
|
||||
}
|
||||
|
||||
func (p *processor) requeue(msg *base.TaskMessage) {
|
||||
err := p.rdb.Requeue(msg)
|
||||
err := p.broker.Requeue(msg)
|
||||
if err != nil {
|
||||
p.logger.Error("Could not push task id=%s back to queue: %v", msg.ID, err)
|
||||
}
|
||||
}
|
||||
|
||||
func (p *processor) markAsDone(msg *base.TaskMessage) {
|
||||
err := p.rdb.Done(msg)
|
||||
err := p.broker.Done(msg)
|
||||
if err != nil {
|
||||
errMsg := fmt.Sprintf("Could not remove task id=%s from %q", msg.ID, base.InProgressQueue)
|
||||
p.logger.Warn("%s; Will retry syncing", errMsg)
|
||||
p.syncRequestCh <- &syncRequest{
|
||||
fn: func() error {
|
||||
return p.rdb.Done(msg)
|
||||
return p.broker.Done(msg)
|
||||
},
|
||||
errMsg: errMsg,
|
||||
}
|
||||
@ -260,13 +260,13 @@ func (p *processor) markAsDone(msg *base.TaskMessage) {
|
||||
func (p *processor) retry(msg *base.TaskMessage, e error) {
|
||||
d := p.retryDelayFunc(msg.Retried, e, NewTask(msg.Type, msg.Payload))
|
||||
retryAt := time.Now().Add(d)
|
||||
err := p.rdb.Retry(msg, retryAt, e.Error())
|
||||
err := p.broker.Retry(msg, retryAt, e.Error())
|
||||
if err != nil {
|
||||
errMsg := fmt.Sprintf("Could not move task id=%s from %q to %q", msg.ID, base.InProgressQueue, base.RetryQueue)
|
||||
p.logger.Warn("%s; Will retry syncing", errMsg)
|
||||
p.syncRequestCh <- &syncRequest{
|
||||
fn: func() error {
|
||||
return p.rdb.Retry(msg, retryAt, e.Error())
|
||||
return p.broker.Retry(msg, retryAt, e.Error())
|
||||
},
|
||||
errMsg: errMsg,
|
||||
}
|
||||
@ -275,13 +275,13 @@ func (p *processor) retry(msg *base.TaskMessage, e error) {
|
||||
|
||||
func (p *processor) kill(msg *base.TaskMessage, e error) {
|
||||
p.logger.Warn("Retry exhausted for task id=%s", msg.ID)
|
||||
err := p.rdb.Kill(msg, e.Error())
|
||||
err := p.broker.Kill(msg, e.Error())
|
||||
if err != nil {
|
||||
errMsg := fmt.Sprintf("Could not move task id=%s from %q to %q", msg.ID, base.InProgressQueue, base.DeadQueue)
|
||||
p.logger.Warn("%s; Will retry syncing", errMsg)
|
||||
p.syncRequestCh <- &syncRequest{
|
||||
fn: func() error {
|
||||
return p.rdb.Kill(msg, e.Error())
|
||||
return p.broker.Kill(msg, e.Error())
|
||||
},
|
||||
errMsg: errMsg,
|
||||
}
|
||||
|
@ -71,7 +71,7 @@ func TestProcessorSuccess(t *testing.T) {
|
||||
cancelations := base.NewCancelations()
|
||||
p := newProcessor(newProcessorParams{
|
||||
logger: testLogger,
|
||||
rdb: rdbClient,
|
||||
broker: rdbClient,
|
||||
ss: ss,
|
||||
retryDelayFunc: defaultDelayFunc,
|
||||
syncCh: nil,
|
||||
@ -178,7 +178,7 @@ func TestProcessorRetry(t *testing.T) {
|
||||
cancelations := base.NewCancelations()
|
||||
p := newProcessor(newProcessorParams{
|
||||
logger: testLogger,
|
||||
rdb: rdbClient,
|
||||
broker: rdbClient,
|
||||
ss: ss,
|
||||
retryDelayFunc: delayFunc,
|
||||
syncCh: nil,
|
||||
@ -253,7 +253,7 @@ func TestProcessorQueues(t *testing.T) {
|
||||
ss := base.NewServerState("localhost", 1234, 10, tc.queueCfg, false)
|
||||
p := newProcessor(newProcessorParams{
|
||||
logger: testLogger,
|
||||
rdb: nil,
|
||||
broker: nil,
|
||||
ss: ss,
|
||||
retryDelayFunc: defaultDelayFunc,
|
||||
syncCh: nil,
|
||||
@ -330,7 +330,7 @@ func TestProcessorWithStrictPriority(t *testing.T) {
|
||||
ss := base.NewServerState("localhost", 1234, 1 /* concurrency */, queueCfg, true /*strict*/)
|
||||
p := newProcessor(newProcessorParams{
|
||||
logger: testLogger,
|
||||
rdb: rdbClient,
|
||||
broker: rdbClient,
|
||||
ss: ss,
|
||||
retryDelayFunc: defaultDelayFunc,
|
||||
syncCh: nil,
|
||||
|
10
scheduler.go
10
scheduler.go
@ -7,13 +7,11 @@ package asynq
|
||||
import (
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
"github.com/hibiken/asynq/internal/rdb"
|
||||
)
|
||||
|
||||
type scheduler struct {
|
||||
logger Logger
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
|
||||
// channel to communicate back to the long running "scheduler" goroutine.
|
||||
done chan struct{}
|
||||
@ -25,14 +23,14 @@ type scheduler struct {
|
||||
qnames []string
|
||||
}
|
||||
|
||||
func newScheduler(l Logger, r *rdb.RDB, avgInterval time.Duration, qcfg map[string]int) *scheduler {
|
||||
func newScheduler(l Logger, b broker, avgInterval time.Duration, qcfg map[string]int) *scheduler {
|
||||
var qnames []string
|
||||
for q := range qcfg {
|
||||
qnames = append(qnames, q)
|
||||
}
|
||||
return &scheduler{
|
||||
logger: l,
|
||||
rdb: r,
|
||||
broker: b,
|
||||
done: make(chan struct{}),
|
||||
avgInterval: avgInterval,
|
||||
qnames: qnames,
|
||||
@ -63,7 +61,7 @@ func (s *scheduler) start(wg *sync.WaitGroup) {
|
||||
}
|
||||
|
||||
func (s *scheduler) exec() {
|
||||
if err := s.rdb.CheckAndEnqueue(s.qnames...); err != nil {
|
||||
if err := s.broker.CheckAndEnqueue(s.qnames...); err != nil {
|
||||
s.logger.Error("Could not enqueue scheduled tasks: %v", err)
|
||||
}
|
||||
}
|
||||
|
@ -36,7 +36,7 @@ type Server struct {
|
||||
|
||||
logger Logger
|
||||
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
|
||||
// wait group to wait for all goroutines to finish.
|
||||
wg sync.WaitGroup
|
||||
@ -208,7 +208,7 @@ func NewServer(r RedisConnOpt, cfg Config) *Server {
|
||||
subscriber := newSubscriber(logger, rdb, cancels)
|
||||
processor := newProcessor(newProcessorParams{
|
||||
logger: logger,
|
||||
rdb: rdb,
|
||||
broker: rdb,
|
||||
ss: ss,
|
||||
retryDelayFunc: delayFunc,
|
||||
syncCh: syncCh,
|
||||
@ -219,7 +219,7 @@ func NewServer(r RedisConnOpt, cfg Config) *Server {
|
||||
return &Server{
|
||||
ss: ss,
|
||||
logger: logger,
|
||||
rdb: rdb,
|
||||
broker: rdb,
|
||||
scheduler: scheduler,
|
||||
processor: processor,
|
||||
syncer: syncer,
|
||||
@ -330,7 +330,7 @@ func (srv *Server) Stop() {
|
||||
|
||||
srv.wg.Wait()
|
||||
|
||||
srv.rdb.Close()
|
||||
srv.broker.Close()
|
||||
srv.ss.SetStatus(base.StatusStopped)
|
||||
|
||||
srv.logger.Info("Bye!")
|
||||
|
@ -10,12 +10,11 @@ import (
|
||||
|
||||
"github.com/go-redis/redis/v7"
|
||||
"github.com/hibiken/asynq/internal/base"
|
||||
"github.com/hibiken/asynq/internal/rdb"
|
||||
)
|
||||
|
||||
type subscriber struct {
|
||||
logger Logger
|
||||
rdb *rdb.RDB
|
||||
broker broker
|
||||
|
||||
// channel to communicate back to the long running "subscriber" goroutine.
|
||||
done chan struct{}
|
||||
@ -24,10 +23,10 @@ type subscriber struct {
|
||||
cancelations *base.Cancelations
|
||||
}
|
||||
|
||||
func newSubscriber(l Logger, rdb *rdb.RDB, cancelations *base.Cancelations) *subscriber {
|
||||
func newSubscriber(l Logger, b broker, cancelations *base.Cancelations) *subscriber {
|
||||
return &subscriber{
|
||||
logger: l,
|
||||
rdb: rdb,
|
||||
broker: b,
|
||||
done: make(chan struct{}),
|
||||
cancelations: cancelations,
|
||||
}
|
||||
@ -49,7 +48,7 @@ func (s *subscriber) start(wg *sync.WaitGroup) {
|
||||
)
|
||||
// Try until successfully connect to Redis.
|
||||
for {
|
||||
pubsub, err = s.rdb.CancelationPubSub()
|
||||
pubsub, err = s.broker.CancelationPubSub()
|
||||
if err != nil {
|
||||
s.logger.Error("cannot subscribe to cancelation channel: %v", err)
|
||||
select {
|
||||
|
Loading…
Reference in New Issue
Block a user