Reducing memory usage by Apache and MySql in a 512MB VPS

Installing with the default Apache and MySQL configurations, you end up with far too many idle Apache threads, and MySQL is the largest memory user by far. The default config of these two is maxing out my available 512MB in my VPS server:

/etc/apache2$ ps -eo pmem,pcpu,rss,vsize,args | sort -k 1 -r       

%MEM %CPU   RSS    VSZ COMMAND
 9.5  0.0 49904 720256 /usr/sbin/mysqld
 6.7  0.0 35620 286860 /usr/sbin/apache2 -k start
 6.5  0.0 34452 283524 /usr/sbin/apache2 -k start
 6.2  0.0 32692 283012 /usr/sbin/apache2 -k start
 5.9  0.0 31276 283116 /usr/sbin/apache2 -k start
 5.8  0.0 30896 282652 /usr/sbin/apache2 -k start
 5.0  0.0 26724 282588 /usr/sbin/apache2 -k start
 4.8  0.0 25204 279552 /usr/sbin/apache2 -k start
 4.8  0.0 25200 279552 /usr/sbin/apache2 -k start
 4.7  0.0 25156 279508 /usr/sbin/apache2 -k start
 4.4  0.0 23216 279540 /usr/sbin/apache2 -k start
 2.8  0.0 15136 278400 /usr/sbin/apache2 -k start
 0.7  0.0  3968  90908 sshd: myuser [priv] 
 0.4  0.0  2564  61312 /usr/sbin/sshd -D
 0.4  0.0  2264  33188 init
 0.3  0.0  2060  18128 -bash
...

/etc/apache2$ free
             total       used       free     shared    buffers     cached
Mem:        524288     476496      47792      68220          0     260764
-/+ buffers/cache:     215732     308556
Swap:            0          0          0

 

From tips in this article, I reduced down the Apache threads and now we’re at:

/etc/apache2/mods-enabled$ ps -eo pmem,pcpu,rss,vsize,args | sort -k 1 -r
%MEM %CPU   RSS    VSZ COMMAND
 9.5  0.0 49904 720256 /usr/sbin/mysqld
 2.8  0.0 15132 278400 /usr/sbin/apache2 -k start
 1.1  0.0  6004 278424 /usr/sbin/apache2 -k start
 0.7  0.0  3968  90908 sshd: myuser [priv] 
 0.4  0.0  2564  61312 /usr/sbin/sshd -D
 0.4  0.0  2264  33188 init
 0.3  0.0  2076  18136 -bash

/etc/apache2/mods-enabled$ free
             total       used       free     shared    buffers     cached
Mem:        524288     400244     124044      68220          0     262480
-/+ buffers/cache:     137764     386524
Swap:            0          0          0

Now we’ve got some free space, and definitely not maxing out our available ram. Next up, I’ll take a look at MySQL.

Setting up WordPress on Apache and MySql on a Linux VPS host

I’ve run this blog for the past couple of years on OpenShift Online. I’ve been excited that the new Online v3 is moving to a container based service, but the cost per month is now much more than I wanted to pay. I was about to setup WordPress on AWS in an EC2 instance (if for nothing more than to get some experience playing with EC2), but after posting on Twitter for suggestions, there was a suggestion to check out cheap Virtual Private Server (VPS) offerings. The lowendbox.com site covers many offers from hosting companies offering VPS based services, so I took a look and picked up a 2 CPU core, 512MB, 200GB disk VPS for $2.50 a month. That’s a manageable cost and looks like a comparably price for a low end VPS server.

To get started after provisioning my VPS , I created a new user with sudo access instead of using the default root user created on initial setup:

  • adduser newuser
  • usermod -aG sudo newuser

On my Ubuntu 14.04 minimal server install, apparently even sudo is not yet installed, so per steps here, su’d to root, and then installed sudo:

  • su -
  • apt-get update
  • apt-get install sudo

Installed mysql-server. There’s many guides for installing MySQL, but here’s one as a reference.

  • sudo apt-get update
  • sudo apt-get install mysql-server

During installation I got this error, and the server failed to start during installation:

/var/lib/dpkg/info/mysql-server-5.5.postinst: line 150: logger: command not found
ATTENTION: An error has occured. More info is in the syslog!

From this post here, the fix is to:

  • apt-get --reinstall install bsdutils

As this ran, the mysql install started showing additional input and the server started up.

Next, setup:

  • sudo mysql_secure_installation

Followed the prompts and remove anonymous user access, test database etc.

Created a new MySQL database with user/password that WordPress will use to access the database, following steps here.

Installed apache:

sudo apt-get install apache2

Installed php modules for apache:

sudo apt-get install php5 libapache2-mod-php5

Hitting the WordPress setup url, got this error:

Your PHP installation appears to be missing the MySQL extension which is required by WordPress.

This is fixed by installing the php mysql package (discussed here):

apt-get install php5-mysql

Installing WordPress from scratch is covered in detail here. The only additional step I needed to do was to tell Apache to server index.php so you’d see the WordPress site when hitting the site root url.

This can be done by editing /etc/apache2/apache2.conf and adding a DirectoryIndex:

<Directory /var/www/>
        Options Indexes FollowSymLinks
        AllowOverride None
        DirectoryIndex index.php
        Require all granted
</Directory>

Restart apache and you should be up and running:

sudo service apache2 restart

As  start this is pretty good. It looks like the default apache and mysql settings are maxing out my 512MB so I’ve got some tuning to do, and then I need to migrate my WordPress database across, but so far so good!

Running Consul for service discovery in Docker containers

I’ve been looking at Spring Boot services in Docker and options for Service Discovery. I started looking at Eureka a while back, but just took a look at Consul.

Here’s my docker-compose.yml that I’ve got working so far. I realize this is just Consul running in 3 containers (1 server, and 2 containers with Consul agents), but I was interested in taking a look at how the containers register with the main server and how it’s configured.

One of the agent containers is running the UI on port 8500, this looks interesting to get an overview of what’s registered with the Consul server.

More to come later.

Never assume you know how something works by only observing its external behavior

As a developer, you should never assume you understand how something works solely by observing what it does. This is especially true if you are trying to fix something and your only understanding of the issue is only the behavior that you can observe.

While you don’t have to understand how something works in order to use it, if you’re trying to fix something, especially software, it helps to understand how something works. The reason is what you observe externally as a problem is usually only a symptom of the problem; it’s rarely the actual problem itself.

Let me give you an extremely simplified example. Let’s say you have an electric car, but you’ve no idea how the electric motor drivetrain works, you just know you press the accelerator pedal and it goes. One morning you get in the car and press the pedal and nothing happens. In diagnosing the issue, the only thing you consider is the external symptoms that you can see: you press the pedal and it doesn’t go. An extremely naive conclusion you could make is that the accelerator pedal is broken (!). So you replace the pedal, but then you’re surprised to find that it still doesn’t work (ok, so this is a contrived example to make the point – if you know enough to be able to replace the accelerator pedal, you probably know enough about how the car works to not assume the pedal is broken!)

As a software developer or architect, as you diagnose issues you should always look under the covers and find out more about what’s actually going on. The problem you’re looking for is rarely the symptom that you can actually see (or what the user sees).