Updated bash testing suite

As a followup to last week's testing post I wanted to expand from a test script, to a testing suite.

Anyone familiar with TDD will often see something like a `tests` directory and a simple `run_tests` script. I wanted to do this in bash.

Using the same style trapping as before, with a simple change, we can achieve a nice result.

#!/usr/local/bin/bash

set -o pipefail

# Begin tests

function error_exit() {

  echo -e "\E[31mTest $TEST failed\E[0m"

  exit 1

}

function success_exit() {

  echo -e "\E[32m $TEST OK\E[0m"

}

trap "error_exit" ERR

trap "success_exit" RETURN

for i in `ls tests/*`; do

  source $i
done

Any file in the tests directory will be evaluated and the results will be reported through the run_tests script

Here's some sample tests:

ls tests/

01-file.sh 02-cows

looking at tests/01-file.sh

export TEST="File exists"

find . -name file &>/dev/null

looking at tests/02-cows:

export TEST="File has cows"

grep -q cows file

Super simple, no shebang interpreter needed, not even an executable file. All we need is to set the value of $TEST to something meaningful. Here's what the test run looks like:

 $ ./run_tests.sh 

 File exists OK

Test File has cows failed

Lets see what's wrong with this...

cat file 

Cows

Ah... wrong case, lets fix that:

 cat file 

cows

$ ./run_tests.sh 

 File exists OK

 File has cows OK

Clean assertion tests in bash

You've likely seen test tools like spec, rspec, and the like, where the tests are written like:

describe package('httpd') do it { should be_installed } end

Recently, I had the need to test guest images using simple assertion tests, like checking for options in sshd_config, or if a package was installed. I initially looked at serverspec which looks like a nice tool, but it seemed to require that my image be a running instance for an ssh session to run the tests.

I figured I could check all these things in a bash chroot environment.

Previously, when running tests, I'd always check for exit status, for example:

grep -q UTC /etc/localtime

if [ $? -ne 0 ]; then

  echo "Failed"

  exit 1

else

  echo "OK"

fi

This seemed like a really tedious thing to keep up as the number of tests increased. Thankfully bash is more and more powerful than most people give it credit for.

Enter traps

I created a generic function for handling errors and successes of tests, and use `trap` to execute those functions based on the signals received. This makes writing tests very easy. More tests means better end products.

function error_exit() {

  echo -e "\E[31mTest $TEST failed\E[0m"

  exit 1

}

function success_exit() {

  echo -e "\E[32m $TEST OK\E[0m"

}

trap "error_exit" ERR

trap "success_exit" SIGCHLD

TEST="Check for UTC timezone"

grep -q UTC /etc/localtime

TEST="Check for cloud-user"

grep -q cloud-user /etc/cloud/cloud.cfg

TEST="Check for no password auth"

grep PasswordAuthentication /etc/ssh/sshd_config | grep -qi no

TEST="Check for puppet client"

which puppet

That's it! Simple. I just set the value of $TEST to a nice descriptive name, and let the traps figure out if the test succeeded or failed. Color output added for a nice extra touch.

Ceph cluster in Vagrant

Working on creating a 3-node ceph cluster in Vagrant...
This allows for a real quick-and-dirty development environment that can be used either 'as-is' PoC, or integrated with devstack.

https://github.com/alop/vagrant-ceph