Chris MacNaughton bio photo

Chris MacNaughton

Chris MacNaughton currently works on the Ecosystem Engineering team at Canonical Ltd, focusing on storage technologies for OpenStack deployments.

Email Public Key Twitter Facebook Google+ LinkedIn Github Stackoverflow

I wanted to add a couple of actions to a Juju charm, specifically Ceph, and I wanted to be able to test the action to ensure that Ceph received the command and modified its settings.

Adding an Action

Adding an action to a charm that currently does not have any actions requires two things, the new action to be written and an actions.yaml be added to the charm’s root directory to describe the new action as well as any options that it takes.

An action can be as sinple as a bash one liner or as complex as a compiled binary, just like a standard Juju hook. I went about adding my two hooks, pause and resume, as simple bash files that tell ceph to pause and resume the Ceph cluster’s health monitoring.

Pause

#!/bin/bash

set -eux

ceph osd set nodown
ceph osd set noout

Resume

#!/bin/bash

set -eux

ceph osd unset nodown
ceph osd unset noout

After adding in my two actions’ executables, I created actions.yaml:

pause:
  description: Pause ceph health operations
resume:
  description: Resume ceph health operations

At this point, I could deploy the Ceph charm as well as run these actions on the cluster. My work was done, right? Wrong! I needed to have tests validating that these actions work; I’d hate to have somebody accidentally break these features at a later date and not be alerted to the break by their tests!

Testing

Adding a new test to an Open Stack charm seems pretty reasonable, add a function into the right place and it’ll run:

def test_402_pause_resume_actions(self):
    """Veryfy that pause/resume works"""
    u.log.debug("Testing pause")
    cmd = "ceph -s"

    sentry_unit = self.ceph0_sentry

Now I’ve got a running test but how do I actually run and verify the actions? After a lot of digging around, I found another Open Stack charm that does test their actions, Swift-proxy, that I could model my test on.

First, a bit of background. The Open Stack charms tend to test a lot of the Open Stack deployment to verify each service. They each define u = OpenStackAmuletUtils(DEBUG) in the setup of the tests. u gives a lot of support methods for the Open Stack charms, including logging and authenticating to the various Open Stack services.

Another thing that the u offers is the ability to run actions :smile: The code I needed was:

action_id = u.run_action(sentry_unit, 'pause')
assert u.wait_on_action(action_id), "Pause action failed."

After doing this, the action had been run on the first Ceph unit in the cluster. After running the hook, I could add:

output, code = sentry_unit.run(cmd)
import re
if re.match('flags nodown,noout', output) is None:
    amulet.raise_status(amulet.FAIL, msg="Missing noout,nodown")

to run ceph -s on the node and check for the expected flags. Testing the resume action is nearly the same, and the whole function is posted below for clarity:

def test_402_pause_resume_actions(self):
    """Veryfy that pause/resume works"""
    u.log.debug("Testing pause")
    cmd = "ceph -s"

    sentry_unit = self.ceph0_sentry
    # action_id = self.action_do(sentry_unit, 'pause')
    action_id = u.run_action(sentry_unit, 'pause')
    # self.action_fetch(action_id)
    assert u.wait_on_action(action_id), "Pause action failed."

    output, code = sentry_unit.run(cmd)
    import re
    if re.match('flags nodown,noout', output) is None:
        amulet.raise_status(amulet.FAIL, msg="Missing noout,nodown")

    u.log.debug("Testing resume")
    action_id = u.run_action(sentry_unit, 'resume')
    assert u.wait_on_action(action_id), "Resume action failed."
    output, code = sentry_unit.run(cmd)
    if re.match('flags nodown,noout', output) is not None:
        amulet.raise_status(amulet.FAIL, msg="Still has noout,nodown")