Overview

kettle is a simple library that abstracts the use of distributed locking to elect a master among group of workers at a specified time interval. The elected master will then call the "master" function. This library uses Redis as the default distributed locker.

How it works

All workers that share the same name will attempt to grab a Redis lock to become the master. A provided master function will be executed by the node that successfully grabbed the lock. A single node works as well, in which case, that node will run both as master and a worker.

The main changes in v2.x.x is the use of context for termination and an optional 'done' channel for notification. It looks something like this:

name := "kettle-example"
k, _ := kettle.New(kettle.WithName(name), kettle.WithVerbose(true))
in := kettle.StartInput{
    // Our master callback function.
    Master: func(v interface{}) error {
        kt := v.(*kettle.Kettle)
        log.Println("from master, name:", kt.Name())
        return nil
    },
    MasterCtx: k, // arbitrary data that is passed to master function
}

ctx, cancel := context.WithCancel(context.TODO())
done := make(chan error, 1)
err = k.Start(ctx, &in, done)
_ = err

// Simulate work
time.Sleep(time.Second * 5)
cancel() // terminate
<-done   // wait

For version 0.x.x, it looks something like this:

name := "kettle-example"
k, _ := kettle.New(kettle.WithName(name), kettle.WithVerbose(true))
in := kettle.StartInput{
    // Our master callback function.
    Master: func(v interface{}) error {
        kt := v.(*kettle.Kettle)
        log.Println("from master, name:", kt.Name())
        return nil
    },
    MasterCtx: k, // arbitrary data that is passed to master function
    Quit:      make(chan error),
    Done:      make(chan error),
}

err = k.Start(&in)
_ = err

// Simulate work
time.Sleep(time.Second * 5)
in.Quit <- nil // terminate
<-in.Done      // wait

Environment variables

# Required
REDIS_HOST=1.2.3.4:6379

# Optional
REDIS_PASSWORD=***
REDIS_TIMEOUT_SECONDS=5

Example

A simple example is provided here for reference. Try running it simultaneously on multiple nodes. For the version 0.x.x example, check it out here.