Implementation of worker pool with extendable functionality.
By default, I've implemented "round-robin" balancing to evenly distribute jobs across workers.
You need to implement the BalancingStrategy
interface in order to use other balancing strategies:
type BalancingStrategy interface {
NextWorkerId(workersStats []Stats) int
}
API:
// worker pool struct private api
(wp *WorkerPool) getCurrentJobsNumber() int64
(wp *WorkerPool) getWorkerStats(workerId int) (Stats, error)
(wp *WorkerPool) terminateWorker(workerId int) error
(wp *WorkerPool) reloadWorker(workerId int) error
// worker pool public methods
// New creates new instance of workers pool
// params defining it's size and max jobs per worker (without blocking)
NewWorkerPool(nWorkers, maxJobs uint, balancer BalancingStrategy) *WorkerPool
// ScheduleJob puts new job in some worker queue
(wp *WorkerPool) ScheduleJob(f JobFunc) (chan Result, error)
// NewRoundRobin creates new instance of the "load balancer"
NewRoundRobin() *RoundRobin
// NextWorkerId returns worker id selected by some
// predefined policy
(rr *RoundRobin) NextWorkerId(workerStats []Stats) int
Install:
go get -u github.com/gasparian/worker-pool-go
Usage example:
package main
import (
"fmt"
wp "github.com/gasparian/worker-pool-go"
)
// Use closure to create job function
func exampleJob(inp int) wp.JobFunc {
return func() wp.Result {
res := inp * inp
return wp.Result{
Data: res,
Err: nil,
}
}
}
func main() {
pool := NewWorkerPool(
3,
10,
NewRoundRobin(),
)
// NOTE: if number of jobs per woker will be > MaxJobs,
// ScheduleJob func will block
nJobs := 50
results := make([]chan wp.Result, 0)
for i := 0; i < nJobs; i++ {
ch := pool.ScheduleJob(exampleJob(i))
results = append(results, ch)
}
for _, r := range results {
res := <-r
fmt.Println(res)
}
}
Run tests:
make test