Hacker News Reader

Comment by zodPod | Hacker News Reader

Kind of comically ironic. I can't currently load the article.. Hug of death I'm guessing? So much for self-hosting?

Keeping the site online despite being self hosted and on the front page of HN might be the ultimate challenge for the people fond of self hosting!

Not using Wordpress to serve static blog content should be a good start.

Also running the database, webserver, mail server, etc. off a single machine is just asking for trouble.

Wordpress can trivially handle "static blog content" with any of a number of caching plugins. (And you get a lot more out of it, in my experience, than trying to use Jekyll or the like.)

nginx, PHP-FPM and APC, your caching tool of choice, done.

And yet after all these years, they've never bothered to roll functionality like that into the base install.

I much prefer, as somebody who uses this stuff pretty heavily, to make the choice of caching particulars myself and to not have to unscrew their defaults. A cache API would be fine, but I'd rather have as few decisions made inside Wordpress as possible.

I know that attitude well, but really this is a problem that's persisted for years and they have basically acted like it's not their problem, which leaves me unimpressed.

Wordpress is fine; this is why things like Varnish exist. If it's static content you can generate it by hand and serve it as fast as possible (after initial generation, obviously).

> Wordpress is fine; this is why things like Varnish exist

Sounds like "now you have two problems"? There is plenty of blog software out there but very few are so terribly, terribly slow as Wordpress. I'm pretty sure it's the slowest thing I've ever installed on my server (and uninstalled, obviously). Just use something else and there is no need for caching in the first place.

Why would the speed of the blog be a problem if it's behind a cache? Just ramp up the ttl entering varnish and purge aggressively on changes; hell, just reboot varnish any time you make a change. Nothing about a cache interface is difficult to work around when your content changes as little as a blog would.

If anything, moving to an unproven blog framework is going to cause much more headaches than a worpress blog slowing down for seconds every time you make a post.

> Also running the database, webserver, mail server, etc. off a single machine is just asking for trouble.

Until five years ago I ran that* on an old 2001 laptop with 128MB RAM, of which about 56MB went to Windows XP. Worked quite fine really.

* That, plus FTP server, SMB, DNS, VNC and uTorrent. I think that was all.

I've been on the front page of HN a couple times, it's not really all that much traffic. Ten thousand pageloads, at most. If your blog can't handle tens of simultaneous connections, you've got problems.

I have the same experience. 10k+ views, maxing out at 100-ish simultaneous. To me a failure to keep the site up is less about the server and more a commentary on the platform the server is running. A single $20/mo box can support a remarkable amount of traffic if you are running lightweight blogging platforms like Ghost.

A single $20/mo box can handle thousands of concurrent hits to a Wordpress site when that site is properly configured. This is user error.

I did that on a 100KB/s uplink and an EEE box as server. And yes, the self-written software loaded the blog from a database (plus tag cloud and stuff) on every page load. The secret (buy my new book!) is to Not Use Wordpress®.

I guess it depends on your point of view. I also self-host, but I do it because I want control of my data. I'm not the blogging type, but even if I were, I would not consider it a failure that my self-hosted blog went down under the weight of hacker news.

It's not like many other professionally hosted websites haven't done the same over the years, and my data remains in my control even when my website is crushed under the weight of enthusiastic visitors, so my self-hosting goal would be intact.

Isn't that... The definition of a failure?

Only if your goal is to support that amount of traffic.

He's saying his goal is controlling his data, not necessarily 100% uptime, so even if the site ends up going down, he still retained control of his data, and therefore did not fail on the metric that is actually important to him.

Maybe if you specialize in hosting large traffic sites, but if the content got enough attention to cause server failure, then it had to be at least moderately successful content.

Looks like it. Here's a mirror: http://webcache.googleusercontent.com/search?q=cache:https:/...

I had the same thought.

Reply on news.ycombinator.com