Dear Ansiblists,
I need to integrate application-level tests with ansible runs. These test are quite specific to that applications we are deploying and go quite beyond the existing test support in ansible. The purpose of these tests is not to test the ansible code but the entire deployment stack. Running these tests as part as part of the ansible playbook makes sense as the ansible run knows the location of all created resources. I am happy to go off and create this on my own, but I would really like to collaborate if there are others working on similar problems. I want to write the tests using plain python unittest so it will be familiar to my developers.
here is an example where i want to test a Consul cluster of 5 nodes. I want to verify that the Consul API functions as expected.
here is a code example of what I am looking to test
tests/test_consul.py
import unittest
import os.environ
import requests
import consul
import time
class TestConsul(unittest.TestCase):
def setUp(self):
self.members = os.environ[‘CONSUL_CLUSTER’].split(‘,’)
def test_nodes_return_same_membership_list(self):
results =
for member in self.members:
member_list = requests.get(‘http://member:4444/v1/cluster/members’ % member)
member_list.sort()
results.append(member_list)
self.assertTrue(all([True for result in results if result == results[0]))
def test_kv_put(self):
c = consul.Consul(self.members[0])
c.kv.put(‘foo’, ‘bar’)
wait for replication
time.sleep(5)
results =
for member in self.members:
c = consul.Consul(member)
results.append(c.kv.get(‘foo’))
self.assertTrue(all([True for result in results if result == ‘bar’]))
as you can see, I plan on passing the list of nodes as an environment variable, but I would love to hear alternate approaches.
I currently plan to trigger the running of these tests by an environment variable.
$ RUN_TESTS=TRUE ansible-playbook my-plays.yml
- name: run tests
when: “{{ lookoup(‘ENV’, ‘RUN_TESTS’)|bool}}”
roles: - execute-tests # executes tests in tests/test_*.py using unittest module, yet to be written
It’s really key that my developers be able to write tests using plain python code rather than Ansible’s YAML. This decouples the tests from ansible. AThis could we helpful later if we every choose to run the exact same tests later separately from ansible, for example executing them as part of regular monitoring.
Thoughts?
Regards,
Bryan W. Berry