FWIW, I went another way here - with apache. Rather than pegging the container to apache running, I used Kubernetes to health check the app, and simply catted out the log files to stdio. It's not really the "docker-way" of one process per container, but it's working pretty well.
The idea was that, though in one container, there would be just one process, I'd run 50 containers on the k8s cluster, and let the cluster take care of health management etc. That would take care of the parallelization issues. Thank you all for the help! On Mon, Oct 26, 2015 at 3:03 PM, Nick Kew <n...@webthing.com> wrote: > On Mon, 26 Oct 2015 07:33:30 -0700 > Kurtis Rader <kra...@skepticism.us> wrote: > > > I don't understand what the problem is. If you're starting a > single-program > > to do exactly one task (e.g., server individual HTTP requests serially) > > That's an unusual use for a general-purpose webserver. > > However, it's perfectly possible. But rather than run in a mode > intended for debug, why not tailor your configuration? > You can tell it to start a single server and limit the maximum > number of servers to one using standard MPM configuration. > > Note that if you're serving regular web pages, the browser > will send you a big flurry of requests for all the images, > scripts, stylesheets, etc in a page. And most browsers will > send those in parallel on several separate connections. > The server normally handles those in parallel, but if you > limit it to a single process and a single thread within that > (I'm not clear if you meant that too), you will slow page > loads by killing parallelism. > > -- > Nick Kew > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > For additional commands, e-mail: users-h...@httpd.apache.org > >
