24 Jul
2014
24 Jul
'14
2:43 a.m.
Hi Paul,
On Wed, Jul 23, 2014 at 11:29:36PM +0100, Paul Tansom wrote:
This time round stopping and restarting Apache has
made no
difference as when I restart the load increases within a few
minutes. Clearly a higher spec VM would help, but it seems silly
throwing resources at it rather than sorting it properly.
The very first thing you should do is make it so that large numbers
of simultaneous HTTP requests cannot bring down your entire VM.
That is most likely happening because too many Apache children are
running at once, using all the RAM and sending the VM into swap.
If so then the first step would be to work out how many Apache
children can actually be safely run simultaneously, and make sure
that MaxClients is set so this maximum value. When the limit is
reached the connections will stack up and your website will cease to
respond, but your VM won't cease to respond.
That will give you some chance to try some other approaches and see
what works best.
Further steps that don't involve increasing the RAM available might
include lighttpd and/or nginx. At some point you're going to need
something capable of interpreting PHP but if you can make as many
requests as possible avoid getting that far, serving things out of
cache, sending static requests to lighttd/nginx etc., then you get
further.
At the very least you can partition the sluggishness to just the
requests that need to hit PHP: you could have PHP being called as CGI
via a fixed number of FastCGI backend servers, while every other
request is served directly by lighttpd/nginx much faster, even under
load.
Of course, Apache is easy, everything works with Apache, all the
docs are for Apache. But Apache is a memory hog. It is a trade-off
to use more memory and have an easier life. :)
Cheers,
Andy
--
http://bitfolk.com/ -- No-nonsense VPS hosting