Ansible is a great tool for configuration management but because of the way it’s designed a common complaint is that it’s not as fast as other tools like Salt, Chef or Puppet. This is because Ansible doesn’t have an agent that listens (although it can) on a host and uses a different type of deployment methodology that is based on SSH. This post isn’t about the pros and cons of each tool, but rather about ways to improve upon Ansible’s default configuration values. By default Ansible ships with very conservative default values. This is smart in my opinion because it offers greater compatibility out-of-the-box. Here I highlight some safe adjustments that can be made to the default configuration for improved performance (speed!)

Real World Playbook Test

For this test I’m using a real-world playbook that I use in my homelab when provisioning a new CentOS VM. It configures some basic things (hostname, ssh keys, etc), installs common packages/utilities and tunes some OS configurations.

I wrote a very simple little playbook for updating my local DNS records for my piholes. For me it’s easier than manually sshing onto each node and editing a file and restarting the service. Here’s the playbook:

I’m limited to about 80MB/s on downloads on my VPC at Digital Ocean, but I run Nzbget for downloading large files from usenet. It doesn’t take long to download at all, but out of curiosity I wanted to see if I could parallelize this and download multiple files at the same. I use Sonarr for searching usenet for freely distributable training videos which then sends them to NZBget for downloading. Since Sonarr can send multiple files to nzbget which get queued up, I figured I can reduce the queue by downloading them at the same time.

Using Ansible and Terraform (devops automation tools), I can spin up VPC on demand, provision them, configure them as nzbget download nodes and then destroy the instances when complete.

I’m limited to about 40MB/s on downloads on my VPC at Digital Ocean, but I run Sabnzbd for downloading large files from usenet. It doesn’t take long to download at all, but out of curiosity I wanted to see if I could parallelize this and download multiple files at the same. I use Sonarr for searching usenet for freely distributable training videos which then sends them to SABnzbd for downloading. Since Sonarr can send multiple files to sabnzbd which get queued up, I figured I can reduce the queue by downloading them at the same time.

Using Ansible and Terraform (devops automation tools), I can spin up VPC on demand, provision them, configure them as sabnzbd download nodes and then destroy the instances when complete.

The instances all run the same sabnzbd config and the instances use haproxy for round-robin distribution. I will probably change this to Consul, but I just wanted something quick so I used a basic haproxy config.