A simple, efficient, concurrent task runner.
- Simple. Run tasks and schedule jobs with GO.
- Database agnostic. Stepper supports MongoDB, Postgresql (beta).
- Concurrent. Stepper can be used in an unlimited number of instances.
- Scalable. Split one task into small subtasks which will run on different nodes.
go get github.com/matroskin13/stepper
go get github.com/matroskin13/stepper/engines/mongo
or
go get github.com/matroskin13/stepper/engines/pg
package main
import (
"log"
"github.com/matroskin13/stepper"
"github.com/matroskin13/stepper/engines/mongo"
)
func main() {
mongoEngine, err := mongo.NewMongo("mongodb://localhost:27017", "example_database")
if err != nil {
log.Fatal(err)
}
ctx := context.Background()
service := stepper.NewService(mongoEngine)
// Will publish a task on startup
if err := service.Publish(ctx, "example-task", []byte("Hello world")); err != nil {
log.Fatal(err)
}
service.TaskHandler("example-task", func(ctx stepper.Context, data []byte) error {
fmt.Println(string(data))
return nil
})
if err := service.Listen(ctx); err != nil {
log.Fatal(err)
}
}
Or you can use PostgresQL:
import "github.com/matroskin13/stepper/engines/pg"
engine, err := pg.NewPG("postgres://postgres:test@localhost:5432/postgres")
if err != nil {
log.Fatal(err)
}
If you use the stepper you will use a lot of things but first of all you will publish and execute tasks. Let's discuss how you can publish tasks.
service.Publish(context.Background(), "example-task", []byte("hello"))
The example shows the simple way to publish a task. The code will publish a task with a name example-task and content hello.
But also the stepper allows you to use additional options.
If you don't want to execute a task immediately you can set up a delay.
service.Publish(
context.Background(),
"example-task",
[]byte("hello"),
stepper.SetDelay(time.Minute * 1),
)
Or you can use particular a date
service.Publish(
context.Background(),
"example-task",
[]byte("hello"),
stepper.LaunchAt(time.Now().Add(time.Minute * 10)),
)
The second part of the Stepper is execution of tasks in queue.
s.TaskHandler("example-task", func(ctx stepper.Context, data []byte) error {
fmt.Println(string(data))
return nil
})
The example shows the simple way to execute a task.
If your handler returns an error, a task will be returned to the queue. And the task will be held in the queue for 10 seconds. But you can set up a delay manually.
s.TaskHandler("example-task", func(ctx stepper.Context, data []byte) error {
ctx.SetRetryAfter(time.Minute) // will be returned in 1 minute
return fmt.Errorf("some error")
})
If you have a log running task, you can bind a state of task (cursor for example), and if your task failed you will be able to continue the task with the last state
s.TaskHandler("example-task", func(ctx stepper.Context, data []byte) error {
var lastId int
if err := ctx.BindState(&lastId); err != nil {
return err
}
iter := getSomethingFromId(lastId) // something like a mongodb iterator or anything else
for iter.Next() {
lastId = ... // do something
if err := ctx.SetState(lastId); err != nil {
return err
}
}
return nil
})
The most powerful feature of the stepper is creating subtasks. The feature allows you to split a long-running task into separate tasks which will run on different nodes. And when all subtasks will be completed the stepper will call a onFinish
hook of parent task.
The following example shows how to spawn subtasks within a main task.
s.TaskHandler("task-with-threads", func(ctx stepper.Context, data []byte) error {
fmt.Println("have received the word for splitting: ", string(data))
for _, symbol := range strings.Split(string(data), "") {
ctx.CreateSubtask(stepper.CreateTask{
Data: []byte(symbol),
})
}
return nil
}).Subtask(func(ctx stepper.Context, data []byte) error {
fmt.Printf("[letter-subtask]: have received symbol: %s\r\n", data)
return nil
}).OnFinish(func(ctx stepper.Context, data []byte) error {
fmt.Println("subtasks are over")
return nil
})
Or you can use existing a subtask:
ctx.CreateSubtask(stepper.CreateTask{
Name: "some-task",
Data: []byte(symbol),
})
If you want to run repeatead task (cron) you can use jobs
s.RegisterJob(context.Background(), &stepper.JobConfig{
Name: "log-job",
Pattern: "@every 10s",
}, func(ctx stepper.Context) error {
fmt.Println("wake up the log-job")
return nil
})
Read https://pkg.go.dev/github.com/robfig/cron#hdr-CRON_Expression_Format for more information about a pattern.
Also you can create subtasks from a job:
s.RegisterJob(context.Background(), &stepper.JobConfig{
Name: "log-job",
Pattern: "@every 10s",
}, func(ctx stepper.Context) error {
fmt.Println("wake up the log-job")
ctx.CreateSubtask(stepper.CreateTask{
Name: "log-subtask",
Data: []byte("Hello 1 subtask"),
})
return nil
}).OnFinish(func(ctx stepper.Context, data []byte) error {
fmt.Println("success job log-job")
return nil
})
The retry middleware allows you to limit a number of retries.
service := stepper.NewService(db)
s.UseMiddleware(middlewares.Retry(middlewares.RetryOptions{
Interval: time.Second * 5,
MaxRetries: 3,
}))
service := stepper.NewService(db)
prometheusMiddleware := middlewares.NewPrometheus()
s.UseMiddleware(prometheusMiddleware.GetMiddleware())
go func() {
http.ListenAndServe(":3999", promhttp.HandlerFor(prometheusMiddleware.GetRegistry(), promhttp.HandlerOpts{}))
}()
if err := s.Listen(context.Background()); err != nil {
log.Fatal(err)
}
The prometheus middleware provides following metrics:
prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "stepper_task_execution",
Help: "Count of all task executions",
}, []string{"task", "status"})
prometheus.NewHistogramVec(prometheus.HistogramOpts{
Name: "stepper_task_duration_seconds",
Help: "Duration of all executions",
Buckets: []float64{.025, .05, .1, .25, .5, 1, 2.5, 5, 10, 20, 30},
}, []string{"task", "status"})