runutil

Documentation

Overview

Package runutil provides helpers to advanced function scheduling control like repeat or retry.

It's very often the case when you need to excutes some code every fixed intervals or have it retried automatically. To make it reliably with proper timeout, you need to carefully arrange some boilerplate for this. Below function does it for you.

For repeat executes, use Repeat:

err := runutil.Repeat(10*time.Second, stopc, func() error {
	// ...
})

Retry starts executing closure function f until no error is returned from f:

err := runutil.Retry(10*time.Second, stopc, func() error {
	// ...
})

For logging an error on each f error, use RetryWithLog:

err := runutil.RetryWithLog(logger, 10*time.Second, stopc, func() error {
	// ...
})

Another use case for runutil package is when you want to close a `Closer` interface. As we all know, we should close all implements of `Closer`, such as *os.File. Commonly we will use:

defer closer.Close()

The problem is that Close() usually can return important error e.g for os.File the actual file flush might happen (and fail) on `Close` method. It's important to *always* check error. Thanos provides utility functions to log every error like those, allowing to put them in convenient `defer`:

defer runutil.CloseWithLogOnErr(logger, closer, "log format message")

For capturing error, use CloseWithErrCapture:

var err error
defer runutil.CloseWithErrCapture(logger, &err, closer, "log format message")

// ...

If Close() returns error, err will capture it and return by argument.

Index

• func CloseWithErrCapture(logger log.Logger, err *error, closer io.Closer, format string, ...)
• func CloseWithLogOnErr(logger log.Logger, closer io.Closer, format string, a ...interface{})
• func Repeat(interval time.Duration, stopc <-chan struct{}, f func() error) error
• func Retry(interval time.Duration, stopc <-chan struct{}, f func() error) error
• func RetryWithLog(logger log.Logger, interval time.Duration, stopc <-chan struct{}, ...) error

Examples

• Repeat
• Retry

Functions

func CloseWithErrCapture

func CloseWithErrCapture(logger log.Logger, err *error, closer io.Closer, format string, a ...interface{})

CloseWithErrCapture runs function and on error tries to return error by argument. If error is already there we assume that error has higher priority and we just log the function error.

func CloseWithLogOnErr

func CloseWithLogOnErr(logger log.Logger, closer io.Closer, format string, a ...interface{})

CloseWithLogOnErr is making sure we log every error, even those from best effort tiny closers.

func Repeat

func Repeat(interval time.Duration, stopc <-chan struct{}, f func() error) error

Repeat executes f every interval seconds until stopc is closed. It executes f once right after being called.

Example

package main

import (
	"context"
	"fmt"
	"log"
	"time"

	"github.com/improbable-eng/thanos/pkg/runutil"
)

func main() {
	// It will stop Repeat 10 seconds later.
	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
	defer cancel()

	// It will print out "Repeat" every 5 seconds.
	err := runutil.Repeat(5*time.Second, ctx.Done(), func() error {
		fmt.Println("Repeat")
		return nil
	})
	if err != nil {
		log.Fatal(err)
	}
}

func Retry

func Retry(interval time.Duration, stopc <-chan struct{}, f func() error) error

Retry executes f every interval seconds until timeout or no error is returned from f.

Example

package main

import (
	"context"
	"errors"
	"fmt"
	"log"
	"time"

	"github.com/improbable-eng/thanos/pkg/runutil"
)

func main() {
	// It will stop Retry 10 seconds later.
	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
	defer cancel()

	// It will print out "Retry" every 5 seconds.
	err := runutil.Retry(5*time.Second, ctx.Done(), func() error {
		fmt.Println("Retry")
		return errors.New("Try to retry")
	})
	if err != nil {
		log.Fatal(err)
	}
}

func RetryWithLog

func RetryWithLog(logger log.Logger, interval time.Duration, stopc <-chan struct{}, f func() error) error

RetryWithLog executes f every interval seconds until timeout or no error is returned from f. It logs an error on each f error.