Hi Urias, On 2010-10-15 at 20:18:11 [+0200], Urias McCullough <umccullough@xxxxxxxxx> wrote: > On Fri, Oct 15, 2010 at 11:11 AM, Urias McCullough > <umccullough@xxxxxxxxx> wrote: > >> http://baron.haiku-os.org/stats?host=vmweb.haiku-os.org&plugin=memory×pan=day&button=Ok > > > > Ok, I'll try to fix this - I should have access to kill it, and it > > should be documented where I think it is :) > > I have bumped the memory on "vmweb" to 2557mb and restarted it (was > killed hard): > > > baron:/ # service VM-web restart > Asking VM web to shutdown (timeout 30 seconds)Connection timed out > during banner exchange > *** failed! > Killing VM web hard! done > Starting VM web done That's fine - once the server is under heavy pressure, it won't serve the ssh login fast enough. The only way to fix that would be to apply some traffic-shaping to give ssh traffic enough room, but I don't really know a lot about how to do that practically ... > Hopefully it will stay up :P so far, it seems to stay in bounds. But, judging from the little info I have right now, the apache processes eat a lot of memory. If at all possible, we should try to move to a threaded apache setup, but I'm not sure that's feasible with Drupal being a php application. Anyway, I suppose we should really invest some more time to optimize our Drupal setup. Ideally, we'd pre-render unchanged pages and serve them statically, at least Drupal seems to support that in principle. cheers, Oliver