样本结构（fabfile）共享通用逻辑并允许多个项目/环境

Question

i am new to python and fabric. we currently use capistrano and have a setup similar to this:

/api-b2b
  - Capfile (with generic deployment/setup info)
  /int - target host config (like ip, access etc.)
  /prod - target host config (like ip, access etc.)
  /dev - target host config (like ip, access etc.)
/api-b2c
  /int
  /prod
  /dev
/application1
  /int
  /prod
  /dev
/application2
  /int
  /prod
  /dev

we are not happy with capistrano for handling our java apps - fabric looks like a better (simpler) alternative.

all the example fabfiles i have seen so far are "relatively simple" in that they only handle one application for different hosts. i'd like to see some code where different app/hosts are handled by the same fabric files/infrastructure (like inheritance etc.) to share the same logic for common tasks like git handling, directory creation, symlinks etc.. i hope you get what i mean. i want the whole logic to be the same, just the apps config is different (git repo, target directory). all the rest is the same accross the apps (same server layout...)

i want to be able to enter something like this

$ cd api-b2b
$ fab env_prod deploy
$ cd api-b2c
$ fab env_prod deploy

or

$ fab env_prod deploy:app=api=b2b
$ fab env_prod deploy:app=api=b2c

any help (and pointers to sample files) highly appreciated

cheers marcel

Answer 1

If you genuinely want reuse amongst your fabric code, the most robust approach is to refactor the commonalities out and make it a python module. There are modules like fabtools and cusine that are good examples of what it is possible to do.

If you're looking to have multiple projects, there are a few ways to achieve that result. Assuming you're using a fabfile directory (rather than a single fabfile.py), you'll have a structure like this.

/fabfile
  __init__.py
  b2b.py
  b2c.py

Assuming that you have:

# b2b.py / b2c.py
from fabric.api import *

@task
def deploy():
    # b2b/b2c logic
    pass

When you run fab -l (with an empty __init__.py ) you'll see:

Available commands:

    b2b.deploy
    b2c.deploy

To get closer to what you're looking for, you can dynamically lookup, from an argument, which deployment target you want to run:

# __init__.py 
from fabric.api import *
import b2b
import b2c

@task
def deploy(api):
    globals()[api].deploy()

Which means that on the command line, I can run fab deploy:api=b2b or fab deploy:api=b2c .

---

Edit 27th Jan

Specifying one or machines for a task to run on can be achieved on the command line with the -H or -R switches, using @task or @role decorators, or the settings in the fabric environment (env.hosts and env.roles). The fabric documentation has extensive examples on the execution model that shows you all the details.

One way to do this (and potentially not the best way depending on your application) is to dynamically alter the host lists based on the api and the target environment.

# __init__.py
from fabric.api import *
import b2b
import b2c

@task
def deploy(api, target='test'):
    func  = globals()[api].deploy
    hosts = globals()[api].deploy_hosts(target)
    execute(func, hosts=hosts) 

And now the b2b.py and b2c.py files will look something like:

# b2b.py / b2c.py
@task
def deploy():
    # b2b/b2c logic
    pass

def deploy_hosts(target):
    return {
        'test' : ['localhost'],
        'prod' : ['localhost'],
        'int'  : ['localhost'],
    }[target]