feat: add durable safe retries with uncertainty tracking and SDK examples
This commit is contained in:
149
sdk/go/examples/retry/main.go
Normal file
149
sdk/go/examples/retry/main.go
Normal file
@@ -0,0 +1,149 @@
|
||||
// retry 演示明确失败的安全重试：同一任务两次尝试，退避由平台持久化管理。
|
||||
package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"errors"
|
||||
"fmt"
|
||||
"os"
|
||||
"strings"
|
||||
"sync/atomic"
|
||||
"time"
|
||||
|
||||
orbit "git.yuntuo.hk/k2safe/orbit/sdk/go"
|
||||
)
|
||||
|
||||
func main() {
|
||||
if err := run(); err != nil {
|
||||
fmt.Fprintln(os.Stderr, err)
|
||||
os.Exit(1)
|
||||
}
|
||||
}
|
||||
func run() error {
|
||||
token, queue := os.Getenv("ORBIT_TOKEN"), os.Getenv("ORBIT_QUEUE_ID")
|
||||
if path := os.Getenv("ORBIT_CREDENTIALS_FILE"); path != "" {
|
||||
raw, err := os.ReadFile(path)
|
||||
if err != nil {
|
||||
return fmt.Errorf("read private credentials: %w", err)
|
||||
}
|
||||
var credentials struct {
|
||||
Token string `json:"token"`
|
||||
QueueID string `json:"queue_id"`
|
||||
}
|
||||
if json.Unmarshal(raw, &credentials) != nil {
|
||||
return errors.New("invalid credentials file")
|
||||
}
|
||||
if token == "" {
|
||||
token = credentials.Token
|
||||
}
|
||||
if queue == "" {
|
||||
queue = credentials.QueueID
|
||||
}
|
||||
}
|
||||
client, err := orbit.NewClient(os.Getenv("ORBIT_URL"), orbit.Options{Token: token})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
operation, err := orbit.NewWorkerMessageID()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
handler := "text.normalize.retry." + operation
|
||||
var calls atomic.Uint32
|
||||
process := func(ctx context.Context, e orbit.WorkerExecution) (json.RawMessage, error) {
|
||||
// 第一次在产生业务效果前明确拒绝，可安全重试；计数仅用于制造示例故障。
|
||||
if calls.Add(1) == 1 {
|
||||
return nil, &orbit.WorkerFailure{Code: "TEMPORARY_REJECTION", Retryable: true}
|
||||
}
|
||||
return normalize(ctx, e)
|
||||
}
|
||||
w, err := orbit.NewWorker(client, orbit.WorkerOptions{QueueID: queue, Capacity: 1, ShutdownTimeout: 5 * time.Second, Handlers: []orbit.WorkerHandler{{Capability: orbit.WorkerCapability{Handler: handler, HandlerVersion: "v1", PayloadVersion: "v1"}, Handle: process}}})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), 20*time.Second)
|
||||
defer cancel()
|
||||
// 示例先受理，再启动消费者；生产通常是独立的长驻 Worker 进程。
|
||||
var tasks []orbit.Task
|
||||
for i, text := range []string{" safe retry "} {
|
||||
payload, _ := json.Marshal(struct {
|
||||
Text string `json:"text"`
|
||||
}{text})
|
||||
accepted, err := client.SubmitTask(ctx, fmt.Sprintf("retry-%s-%d", operation, i), orbit.SubmitTaskRequest{QueueID: queue, Handler: handler, HandlerVersion: "v1", PayloadVersion: "v1", Payload: payload, TaskTimeoutMS: 60000, Retry: orbit.RetryPolicy{MaxAttempts: 3, AttemptTimeoutMS: 10000, InitialDelayMS: 500, MaxDelayMS: 2000}, RecoverySafety: orbit.RecoveryReconcileOnly})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
tasks = append(tasks, accepted.Task)
|
||||
}
|
||||
workerCtx, stop := context.WithCancel(ctx)
|
||||
done := make(chan error, 1)
|
||||
go func() { done <- w.Run(workerCtx) }()
|
||||
// 任一调用失败时也等待有界排空，不在 main 返回时静默丢掉执行中的 Worker。
|
||||
waited := false
|
||||
defer func() {
|
||||
stop()
|
||||
if !waited {
|
||||
<-done
|
||||
}
|
||||
}()
|
||||
for _, task := range tasks {
|
||||
for {
|
||||
current, err := client.GetTask(ctx, task.ID)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if current.State == orbit.TaskSucceeded {
|
||||
history, err := client.ListTaskAttempts(ctx, current.ID, orbit.HistoryPageRequest{})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if current.AttemptCount != 2 || len(history.Items) != 2 || history.Items[1].State != orbit.AttemptFailed || !history.Items[1].Retryable || current.EffectUncertain {
|
||||
return errors.New("retry history mismatch")
|
||||
}
|
||||
fmt.Printf("task=%s state=%s attempts=%d first_retryable=%t result=%s\n", current.ID, current.State, current.AttemptCount, history.Items[1].Retryable, current.Result)
|
||||
break
|
||||
}
|
||||
if current.State == orbit.TaskFailed || current.State == orbit.TaskNeedsReview {
|
||||
return fmt.Errorf("task %s requires attention: %s", current.ID, current.State)
|
||||
}
|
||||
select {
|
||||
case err := <-done:
|
||||
waited = true
|
||||
return fmt.Errorf("Worker stopped before completion: %v", err)
|
||||
case <-ctx.Done():
|
||||
return ctx.Err()
|
||||
case <-time.After(100 * time.Millisecond):
|
||||
}
|
||||
}
|
||||
}
|
||||
stop()
|
||||
err = <-done
|
||||
waited = true
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
stats := w.Stats()
|
||||
if stats.Started != 2 || stats.ReportsConfirmed != 2 || stats.Running != 0 {
|
||||
return errors.New("unexpected Worker completion counts")
|
||||
}
|
||||
fmt.Printf("worker_drained=true started=%d confirmed=%d running=%d\n", stats.Started, stats.ReportsConfirmed, stats.Running)
|
||||
return nil
|
||||
}
|
||||
func normalize(ctx context.Context, execution orbit.WorkerExecution) (json.RawMessage, error) {
|
||||
var in struct {
|
||||
Text string `json:"text"`
|
||||
}
|
||||
if err := json.Unmarshal(execution.Task.Payload, &in); err != nil {
|
||||
return nil, &orbit.WorkerFailure{Code: "INVALID_TEXT"}
|
||||
}
|
||||
// 模拟可取消的短工作；真实 I/O 也必须使用传入的 ctx。
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
return nil, ctx.Err()
|
||||
case <-time.After(200 * time.Millisecond):
|
||||
}
|
||||
return json.Marshal(struct {
|
||||
Text string `json:"text"`
|
||||
}{Text: strings.ToUpper(strings.TrimSpace(in.Text))})
|
||||
}
|
||||
Reference in New Issue
Block a user