TLDR: Running commands at multiple hosts with Fabric

Published on February 17, 2018

Recently someones asked me how to perform a connection to a server to run the command df -h using ssh to display the disk usage of the machine. After a quick chat, I discovered that the question was more than just a single check to a server. It was to extend that to multiple servers.

The answer to the question was quick and came with a “click”. How about to document these quick tips at the blog? It can help someone and also help me to remind me how to do something if I need it in the future. With that in mind, I created this series of posts that I kindly baptized as TLDR the abbreviation for Too Long ** D**idn’t Read.

Now moving to the actual tip, to automate repetitive tasks like at the question above or on more complex processes like on software deployment there is a handy python library called Fabric that we are going to use to help us on that.

To install Fabric you can use pip:

$ pip install Fabric3

Then we create the fabfile.py with the following content:

from fabric.api import env, run

env.hosts = ['user1@server1', 'user2@server2']


def disk_usage():
    run('df -h')

The file must be called fabfile.py so Fabric can automatically discover him when we run the command.

Fabric makes the ssh connection with the servers so that you will need an ssh key configured at your machine and its access allowed at the allowed_hosts file of the target servers.

With the file ready, you need to run this command:

$ fab disk_usage
[user1@server1] Executing task 'disk_usage'
[user1@server1] run: df -h
[user1@server1] out: Filesystem      Size  Used Avail Use% Mounted on
[user1@server1] out: /dev/root        20G  2.2G   17G  12% /
[user1@server1] out: devtmpfs        488M     0  488M   0% /dev
[user1@server1] out: tmpfs           490M     0  490M   0% /dev/shm
[user1@server1] out: tmpfs           490M   51M  440M  11% /run
[user1@server1] out: tmpfs           5.0M     0  5.0M   0% /run/lock
[user1@server1] out: tmpfs           490M     0  490M   0% /sys/fs/cgroup
[user1@server1] out: tmpfs            98M     0   98M   0% /run/user/0
[user1@server1] out:

[user2@server2] Executing task 'disk_usage'
[user2@server2] run: df -h
[user2@server2] out: Filesystem      Size  Used Avail Use% Mounted on
[user2@server2] out: /dev/root        20G  3.9G   15G  22% /
[user2@server2] out: devtmpfs        488M  4.0K  488M   1% /dev
[user2@server2] out: none            4.0K     0  4.0K   0% /sys/fs/cgroup
[user2@server2] out: none             98M  352K   98M   1% /run
[user2@server2] out: none            5.0M     0  5.0M   0% /run/lock
[user2@server2] out: none            490M     0  490M   0% /run/shm
[user2@server2] out: none            100M     0  100M   0% /run/user
[user2@server2] out:


Done.

Fabric will run the task in sequence at the host list informed. If you want to run it in parallel use the -P flag:

$ fab disk_usage -P

Bonus

Tip from @luizirber at twitter.

Fabric has a pattern to run its commands at the CLI without needing to write code:

$ fab [options] -- [shell command]

The disk usage command woud be:

$ fab -H 'user1@server1, user2@server2' -- df -h

This work is licensed under a Creative Commons Attribuition-ShareAlike 4.0 International License .