wang502 / gores Goto Github PK

Inside dispacher.go

func (disp *Dispatcher) Dispatch(wg *sync.WaitGroup){
    for {
        select {
        case workerID := <-workerIDChan:
            go func(workerID string){      // goroutine g1
              for {
                job := ReserveJob(disp.resq, disp.queues, workerID)
                if job != nil {
                  disp.jobChannel<-job
                }
              }
            }(workerID)
        case <-time.After(time.Second * time.Duration(disp.timeout)):
            log.Println("Timeout: Dispatch")
            break
        }
        break
    }
    wg.Done()
}

Goroutine g1 will stop after the Dispatch() function returns. When there is no more workerID fetched from worker channel and time is out waiting for worker channel, the outer for loop breaks and Dispatch() returns, so then g1 will stop.

Since gouroutine g1 is used to fetch jobs from Redis, after it stops, there might be unfinished jobs waiting in Redis queue.

Right now, Dispatcher struct has a jobChan with length equals to the number of workers, workers keeps trying to fetch job from jobChan. The execution time of each worker is different, so it may occur that some workers execute a lot more jobs than other workers.

Solved in commit 57

When my account was compromised a spam issue was created in this repo. I sincerely apologize. Cleaning up such issues via script.

The Job interface is not very necessary. Since the job struct is more about the field inside of it, rather than about the behavior or methods. Just create mock job structs in the job_test.go