Earlier this month I talked about how my server was acting wonky and how I fixed it using, among other tools, TOP.
This week I was chatting with a fellow about CPU usage and his site. He runs a rather large WordPress blog and the database is about 500 megs. As a comparison, this site, with about 500 posts, is under 5 megs, and my big site, with thousands of posts, comments, and a forum, is 10 megs. The biggest site I run on my server is 850 megs (just down from 910 after some clean up). The difference between his site and mine is that his is slow and he knows it. As we discussed ways to speed it up, I had some thoughts on WordPress and how, at a certain point, you’re going to need to dig into the guts of your server and learn TOP.
The ‘problem’ with most ‘How do I make my WordPress site run faster?’ tutorials, as I’ve seen it, is they address surviving the digg effect. That is, they talk about how to deal with having a high volume of traffic on your site and, for the most part, you can make it with just adding caching plugins.
Once your site gets ‘big’ or ‘popular’ you’re going to have to move off shared/cloud hosting and over to your own server. For most of us, the first step is a VPS (Virtual Private Server). Shared Hosting means ‘You have an account on a server with a hundred other people.’ It’s great for small sites, inexpensive and easy to use. The problem is you could have terrible neighbors, who use up all the CPU. Think of it like those old New York apartments where someone’s a jerk at 5am and uses up the hot water so you, at 7am, have none. Yeah, it’s kind of like that. That’s the day you think ‘I want a house!’
Only, well, we’re not all up for houses just yet. A house would be a dedicated server, where it’s just you. Cloud hosting, which I touched on earlier, would be the college dorms of webhosting. It has a lot of benefits for the really small sites, and actually some for large sites, but I’m not sold of their overall usefullness yet, so I’ll talk about them some other time. What I want to talk about are Virtual Private Servers, the condo-sub-leasing (or rent-to-own maybe) of website hosting, and how the new VPS user should really get on TOP of things (sorry, bad pun) to make their lives easier.
TOP. Well ‘top’ really. Unix commands are generally all lower case like that.
The top command is a system monitor tool that outputs a list of processes. Have you ever seen Task Manager in Windows? It’s kind of like that tab for ‘Processes’ that you look at and run away from. The default view of top is by percentage of CPU usage and the “top” CPU users are listed. See? The name made sense. You can also see how much processing power is being used, memory hogs and other cool things. Most modern Unix-systems let you sort the list, colorize it, etc, though you have to be command line savvy.
Here’s what top looked like for me about an hour ago.
top - 12:44:44 up 126 days, 23:13, 1 user, load average: 0.12, 0.17, 0.17 Tasks: 91 total, 1 running, 90 sleeping, 0 stopped, 0 zombie Cpu(s): 0.0% us, 0.0% sy, 0.0% ni, 100.0% id, 0.0% wa, 0.0% hi, 0.0% si Mem: 524288k total, 358248k used, 166040k free, 0k buffers Swap: 0k total, 0k used, 0k free, 0k cached PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 15616 nobody 15 0 94540 65m 20m S 1.0 12.8 0:00.27 httpd 12261 ipstenu 16 0 1908 1012 780 R 0.4 0.2 0:00.30 top [...] 28630 root 16 0 107m 86m 1096 S 0.0 16.8 1:06.30 /usr/sbin/clamd
I wanted to point out clamd, which has been the bane of my existance. Thing won’t DIE. I ended up going in to
/etc/exim.conf and manually commended out the clamd line (and restarted the service) to finally get it gone.
But top, as you can see, has a freakishly large amount of information. My server is doing fine, at this point, so I don’t have a whole lot to show you. What you can see right away is that I can tell, with a glance, what’s going on. I could see, though and at this point I have a ‘nobody’ process. That just means someone’s accessing my website. No, really! That’s good! The CPU and memory usage seem high, but they vanish in a second. Basically, someone rang my doorbell and for that brief moment, electricity was used. The next thing I see is the top command, which is run by me (hi!) and down the line is that idiot, clamd.
I actually scan top a lot at work these days, trying to understand what’s causing issues. It’s good for ‘right now!’ things, but not so much if I want to see what started a strange spike a couple hours (weeks) ago. For that you need a whole mess of tools.