[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] PATCH: 18/25: Dynamic thread workers pool



On Tue, Jan 13, 2009 at 05:46:08PM +0000, Daniel P. Berrange wrote:
> @@ -1948,6 +2000,26 @@ static int qemudRunLoop(struct qemud_ser
>              }
>          }
>  
> +        /* If number of active workers exceeds both the min_workers
> +         * threshold and the number of clients, then kill some
> +         * off */
> +        for (i = 0 ; (i < server->nworkers &&
> +                      server->nactiveworkers > server->nclients &&
> +                      server->nactiveworkers > min_workers) ; i++) {
> +
> +            if (server->workers[i].active &&
> +                !server->workers[i].processing) {
> +                server->workers[i].quit = 1;
> +
> +                virCondBroadcast(&server->job);
> +                virMutexUnlock(&server->lock);
> +                pthread_join(server->workers[i].thread, NULL);
> +                virMutexLock(&server->lock);
> +                server->workers[i].active = 0;
> +                server->nactiveworkers--;
> +            }
> +        }
> +

Doesn't this cause the main loop to hang -- eg. if we happen to try to
kill of a worker which is doing some lengthy operation?

> +struct qemud_worker {
> +    pthread_t thread;
> +    int active :1;
> +    int processing :1;
> +    int quit : 1;

I guess maybe I'm unclear about the meaning of these flags.  What's
the difference between active & processing?

Rich.

-- 
Richard Jones, Emerging Technologies, Red Hat  http://et.redhat.com/~rjones
Read my OCaml programming blog: http://camltastic.blogspot.com/
Fedora now supports 68 OCaml packages (the OPEN alternative to F#)
http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]