r3dux.org

When I migrated this site from shared hosting to a VPS, I initially went with some really cheap hosting, and I got what I paid for: The site went down often, like every week or so. Sometimes for days at a time. Support tickets went ignored and unanswered, and I was generally a sad puppy.

After a month or two of this I’d had enough, and decided to migrate the site hosting away from Cheap-O-Host to a company called Linode – and I’ve got to say – it’s been absolutely flawless ever since. I can’t even remember the last time the site went down.

This isn’t to say I haven’t had issues with Apache bogging down and basically killing the site by trying to use more RAM than the VPS has available. If I got more concurrent connections than I had RAM for, then some il-advised Apache settings meant that the server would get into swap territory, and the entire thing would grind to a halt from which the only escape was to log into the VPS from the web interface and reboot it. But this wasn’t Linode’s fault – it was mine for trying to do to much with too little.

I finally solved the problem by disabling all the apache modules I didn’t need (which lowered the RAM usage per Apache process, as well as lowering the server’s attack surface), and limiting the concurrent connections to a number that would happily fit within 512MB of RAM – and since then, not a cough or a hiccup. Everything has been running beautifully.

In fact, at the time of writing, my VPS uptime stats are:

# uptime
21:15:38 up 244 days, 12:26,  1 user,  load average: 0.10, 0.07, 0.06

244 days without as much as a reboot! Sweet!

I’m just writing this post to commend Linode because I got yet another monthly email (from siteuptime.com) today saying “site up-time: 100%” and… well, if something goes wrong then people (and definitely me included, here) tend to jump up and down and make a noise – but when something goes right – we generally don’t say anything at all. Which makes the cost of failure getting slagged off, and the reward for success merely not getting slagged off.

As that’s not a particularly good system, I wanted to say something positive about Linode, because it’s not an accident that their systems so rarely go down – I’m sure that it’s entirely due to their hard work, professionalism, and investment & re-investment in the company infrastructure. And I absolutely cannot fault them on that.

So well done, Linode! You guys rock! =D

P.S. For anyone trying to minimise their apache footprint, the minimal set of apache modules I found I needed to run a wordpress site are:

• alias
• authz_default
• deflate
• dir
• env
• expires
• headers
• mime
• php5
• reqtimeout
• rewrite, and
• setenvif

And my final, working Apache prefork MPM settings in apache2.conf for a VPS with 512MB of RAM are:

# prefork MPM
# StartServers: number of server processes to start
# MinSpareServers: minimum number of server processes which are kept spare
# MaxSpareServers: maximum number of server processes which are kept spare
# MaxClients: maximum number of server processes allowed to start
# MaxRequestsPerChild: maximum number of requests a server process serves
<IfModule mpm_prefork_module>
    StartServers          4
    MinSpareServers       1
    MaxSpareServers       1
    MaxClients            13
    MaxRequestsPerChild   5000
</IfModule>

In my experience, the above config will happily serve tens of thousands of hits per day without getting to a stage where a VPS with 512MB of RAM starts to page out excessively, and without dropping incoming connections. Of course, this won’t stop a site getting flattened, but it’s a pretty good middle ground to cope with a large amount of hits with a low amount of RAM.

Cheers!

That’s just astonishing… wow!