Finglonger is a small tool that turns an intention log into shell commands that are run by a robot on a server. This means you can use git and code review to poke servers instead of rooters logging in.

Finglonger is a 'no-ops' tool. It provides a simple structure so that system administrators don't have to type commands on servers. Instead, system administrators, devs, and others in the organization write their commands into a simple text format and submit that through a pull request or other code review tool. Administrators merge these pull requests, and finglonger runs the commands.

Configuration management is built on the idea of idempotency. But there are times in system administration where something is too rare, too complicated, or too dangerous to run every 20 minutes. Finglonger lets teams coordinate the running of one off tasks, adding git history and code review to what used to be manual root-shell activities.

Finglonger has a task list. It runs each command in the list exactly once. Finglonger does this by using git to track state.

The finglonger-tasks repository has two branches: master and done. Users propose new tasks to the master branch, administrators merge those pull requests, finglonger picks up changes to the master branch, runs the commands, then merges those changes into the done branch. Finglonger manages the 'done' branch and the administrators manage the 'master' branch. Finglonger operates on the difference between the 'master' branch and the 'done' branch.

You need a repo to hold the log. Make this.

• Clone finglonger into /opt/git/finglonger (or wherever)
• virtualenv /opt/git/finglonger/venv
• /opt/git/finglonger/venv/bin/pip install pyyaml
• Clone the log repo into /opt/git/finglonger-tasks/
• Create /var/run/finglonger.lock and /var/log/finglonger.log such that the user running finglonger can write to them
• (optional) setup logrotate on the finglonger log file
• Set some kind of a cron job to run git pull and finglonger.py in the log git repo periodically. It will automatically run things from the log.

*/5 * * * * cd /opt/git/finglonger-tasks && git pull && flock -n /var/run/finglonger.lock /opt/git/finglonger/venv/bin/python /opt/git/finglonger/finglonger.py >> /var/log/finglonger.log 2>&1

The above crontab would cause finglonger to run once every five minutes and log to /var/log/finglonger.log:

You need a finglonger tasks repo. Start out with:

envs/default/tasks.yaml
files/
scripts/

See https://github.com/nibalizer/finglonger-tasks as an example Tasks will only be run in the environment they are set in.

Start your tasks file (tasks.yaml) out looking like this:

---
- task:
    name: job
    shell: echo test +
- task:
    name: job
    shell: echo test +

Add one task object per commit or things are gonna break for you.

Finglonger is controlled by ~/.config/finglonger/config.yaml

---
environment: myenv

Environments allow different servers to be controlled by different tasks files.

Finglonger is super beta, it will probably set your computer on fire. Try it, break it, and lets make it better together.

Yup. Two big goals here:

1. Don't depend on ansible

2. Don't rewrite ansible

It is more important to accomplish the first than the second.

- task:
    name: Hello world job
    shell: echo 'Hello, World!'

(Add config.js to the files directory in your finglonger-tasks)

- task:
    name: Copy config.js to /var/www/html/myapp on app-server1.example.com
    shell: scp files/config.js [email protected]:/var/www/html/myapp/config.js

- task:
    name: Restart nova-api on compute1
    shell: ssh [email protected] 'service nova-api restart'

- task:
    name: Some task that takes a while
    shell:  |
      echo "starting the big script woo"
      dd if=/dev/zero of=/tmp/bigfile bs=1k count=1000

- task:
    name: Something we care about the status of
    shell:  |
      echo "starting the big rsync woo"
      rsync -PHaze ssh my_dir remotehost:/home/nibz/my_dir | tee /var/www/html/finglonger/rsync_log1.txt