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

Re: Socket activation



On Wed, Sep 4, 2013 at 12:09 PM, Clayton Coleman <ccoleman redhat com> wrote:
> From a paas provider perspective there is more advantage to the provider for individual activation because it allows for different classes of container code to be idled at different rates.  For instance a tool like phpmyadmin or Jenkins, running in a container, would ideally be idled at an extremely aggressive rate, whereas the whole unit   may still be taking web requests.

We actually do this with the out-of-band model, too. We idle PHP-FPM
pools 4-10x faster than MariaDB instances. There just has to be a way
for the edge/application container to know that activation has to
occur. It's possible, for example, to first register the MariaDB
container as idle, then idle it. The edge/application container would
need a very fast way to know whether downstream containers need
activation, though.

> OpenShift's current idling mechanism at the http level only works cleanly as long as the primary application front is also http - as more categories of application become relevant, listening to external ports beyond 80/443 we have to adopt a more generic tcp queueing/unifying mechanism anyway, and so socket activation is ideal in concept.

I agree.

-- 
David Strauss
   | david davidstrauss net
   | +1 512 577 5827 [mobile]


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