chapter05.rst

Chapter 5: Deployment with the Salt Client API

So now we have a large collection of states to manage all our different nodes. We have a collection of modules so we can run specific tasks, now we want to be able to have a very simple deployment strategy.

Sure we could in theory log on to the salt master and start typing out a bunch of salt -G 'roles:webserver' myproject.dosomthing but that would be time consuming and prone to error if you happened to forget a step.

Here is where we can utilize the salt client API. The salt client api is just a very simple way to run salt modules, programatically.

Lets say we want to have a "quick_deploy" (the updated code doesn't make any DB changes, it doesn't make any changes to the installed apps or requirements).

A typical flow for doing a quick deploy could be like this:

1. update the code
2. collectstatic
3. restart apache
4. restart supervisor (assuming your're using supervisor to manage celery)

We can make a very simple python script called quick_deploy.py where we essnetially run all these steps in order:

import salt.client
client = salt.client.LocalClient()

ret = client.cmd(
    'G@roles:webserver or G@roles:redisserver',
    'myproject.update_code',
    [],
    expr_form="compound")
print ret

ret = client.cmd('web1', 'myproject.collectstatic')
print ret

ret = client.cmd(
    'roles:webserver',
    'myproject.restart_apache',
    [],
    expr_form="grain")
print ret

ret = client.cmd(
    'roles:redisserver',
    'myproject.restart_supervisord',
    [],
    expr_form="grain")
print ret

This very simple script does the following steps:

1. It does a "compound" target of multiple roles: 'G@roles:webserver or G@roles:redisserver' and it updates the code.
2. It does a simple "hostname" match on web1 to run collectstatic. Since we are hosting our static media on S3, we only need to run collectstatic once, there's no need to target multiple webservers and repeat the job.
3. It does a "grain" match on the webservers and restarts apache on both.
4. it does a "grain" match on the redisserver and restarts supervisor.