Pull to refresh

Test me if you can. Do YML developers Dream of testing ansible?

Reading time3 min

kitchen-ci schema

It is text version of the presentation 2018-04-25 at Saint-Petersburg Linux User Group. Configuration example locates at https://github.com/ultral/ansible-role-testing

I suppose that that you make configuration management, not bash. It means that you have to test it some how. Have you ever tested ansible roles? How do you do it?

How to do it?

In my case, we have:

  • A lot of different ansible roles.
  • Hyper-V hosts as a hypervisor.
  • A private cloud with limited possibility to create VMs on demand.
  • A proxy for internet access.
  • Inability test ansible roles inside docker, because of a role = whole VM configuration.
  • Decision to implement green build policy for git repository with ansible roles.

Let us compare existing solutions for testing.

Name Test Kitchen Molecule Create new
Language ruby python bash/ruby
Watchers 132 126 0
Stars 1413 1154 1
Forks 502 174 2
License Apache 2.0 MIT Any
Commits 1929 1264 0
Releases 101 121 0
Contributors 109 82 5

Name testinfra serverspec inspec Goss
Github philpep/testinfra mizzy/serverspec chef/inspec aelsabbahy/goss
Language python ruby ruby go
Watchers 93 145 165 67
Stars 997 2105 1167 2170
Forks 138 361 330 156
License Apache 2.0 MIT Apache 2.0 Apache 2.0
Commits 380 1854 4609 309
Releases 35 282 346 47
Contributors 43 110 159 31

We decided not to reinvent the wheel & get production ready solution. Our infrastructure team had strong ruby skills & great experience with ruby, as a result we chose Test Kitchen & inspec


kitchen-ci schema

The main idea is to create a new VM, apply an ansible role & do some smoke tests.

Green Build Policy

Green build policy schema

Also, we implemented green-build policy. We ran tests for each commit into master branch & if tests are ok, when apply ansible roles.

Nested virtualization

As you remember, we had a private cloud with limited possibility to create VMs on demand. We decided to create VMs inside VMs.

we need to go deeper

First of all we tried to run Virtualbox x32 without nested. It was bad idea because of kernel panic. Also vast majority of our VMs in our infrastructure are x86_64, so we decided to continue research. As a result we decided to use nested virtualization. Hopefully it was supported by our host servers.

Faced issues

I was implementing testkitchen and faced with some issues.

Pass proxy settings from host into testkitchen guest VM

In some test suits, we configured proxy client settings inside VM created by testkitchen. However, proxy was not configured on testkitchen host & ansible cannot use extra variables with empty values

Solution: create erb template for setting default proxy if ENV variables is not set

<%= ENV['http_proxy'].to_s.empty? ? 'http://proxy.example.com:3128' : ENV['http_proxy'] %>

Manage network settings via playbook

Some roles configure network interfaces. Test suit was looked like:

  • Deploy network settings to VMs
  • Reload network
  • It is failed

Solution: add interfaces to VMs

Fail if suit cases contains "-" in name

Virtualbox can't use "_" in a vm name

Solution: rename suitcases "vm_" => "vm-"

Oracle test fails without "." at the end of VM name

We use role in production, however when we decided to test it, it was failed. We reproduced it.

I would like to show clue.

[root@vm-oracle vagrant]# getent ahosts vm-oracle STREAM vm-oracle DGRAM RAW
[root@vm-oracle vagrant]# getent ahosts vm-oracle.
fe80::a00:27ff:febd:bd6a STREAM vm-oracle
fe80::a00:27ff:febd:bd6a DGRAM
fe80::a00:27ff:febd:bd6a RAW STREAM DGRAM RAW
[root@oracle vagrant]# getent ahosts oracle.example.com. STREAM oracle.example.local DGRAM RAW

Do you have any ideas what is happening?

It is tricky bug:

  1. we enabled listen IPv4 only in oracle listener settings
  2. oracle used FQDN
  3. linux contains special database "myhostname" for resolving hostname, it is uses after /etc/hosts & dns resolving
  4. Vagrant created VM & updated /etc/hosts

I'd like to clarify it a bit more:
What has happened in case vm-oracle

  1. vagrant created vm 
  2. vagrant updated /etc/hosts(vm-oracle x2)
  3. oracle listener listened IPv4
  4. oracle listeners resolved vm-oracle. & geted IPv6

What has happened in case vm-oracle.

  1. vagrant created vm 
  2. vagrant updated /etc/hosts ( vm-oracle &  vm-oracle.)
  3. oracle listener listened IPv4
  4. oracle listeners resolved vm-oracle. & geted IPv4
  5. OK

OOM is coming

OOM randomly was killing VMs. Testkitchen was failed with strange errors.

Solution: increase RAM

Slow builds

It was working slowly


  • Packer. Prebuilded vagrant box with common tasks
  • Concurrency


On one hand, current implementation works, but on the other hand, there are some issues

  • is not user friendly.
  • we mix ruby & python.
  • there is no indepotence check.
  • it works slow.
  • it is hard to trace logs at single job.

As a result, molecule & docker might be pretty interesting solution.

Some related links

Total votes 18: ↑17 and ↓1+16