A Python library that executes commands in parallel, locally and remotely, over a group of hosts.
This library has been extracted from existing projects where it was used as a basis to create deployment scripts. It's a bit like a mini Ansible, without the YAML overhead, and usable as a simple composable library.
Here are some important facts about deploykit:
- Local commands and user-defined function: In contrast to many other libraries in the space, deploykit allows to also run commands and user-defined functions locally in parallel for each host.
- OpenSSH: To retain compatibility with existing configuration, deploykit uses the openssh executable for running the commands remotely.
- Threaded: Deploykit starts a thread per target host run commands and user-defined functions and collects their results for inspection. To run commands, deploykit wraps around python's subprocess API.
- Clean output: command outputs are prefixed by the hostname of the target.
from deploykit import parse_hosts
import subprocess
hosts = parse_hosts("server1,server2,server3")
runs = hosts.run("uptime", stdout=subprocess.PIPE)
for r in runs:
print(f"The uptime of {r.host.hostname} is {r.result.stdout}")A more comprehensive example explaining all the concepts of the API can be found here.
- Fabric:
- Deploykit took inspiration from fabric and addresses its limitation that local commands cannot be executed in parallel for a range of hosts (i.e. rsync). By also allowing to run a function per host it provides higher flexibility. Fabric uses pyinvoke as a task runner frontend, so deploykit can be also used in combination with the same.
- Ansible:
- Deploykit is more lightweight and has a faster startup time.
- Using python for task definitions allows for more flexibility than YAML.
- Use ansible if you need declarative configuration management. Use deploykit if you want to imperatively quickly execute a series of commands on a number of hosts.
Contributions and discussions are welcome. Please make sure to send a WIP PR or issue before doing large refactors, so your work doesn't get wasted (in case of disagreement). In our development documentation we explain how to get starte on deploykit.
This project is copyright Numtide and contributors, and licensed under the MIT.