Saving 216 Minutes in 2.5 Years with Simple Docker Enhancements

Written by - Millan Kaul

Every second spent waiting for a container build is a second of productivity lost. Recently, I looked at the Docker setup for my own Jekyll blog (this blog you are reading right now) and realized it was taking 96.0 seconds to start up from scratch with docker compose up.

Here is how I optimized the setup, shaving it down to 63.5 seconds (a 33.8% speedup), 😱 and the math behind how much time this saves over the project’s lifetime.

---

The Problem Statement

Our old Docker configuration was slow due to three main issues:

1. Redundant installation: The Dockerfile ran gem install jekyll bundler even though the base ruby image already includes Bundler, and Jekyll is managed by Gemfile.
2. Permission overhead: The build step ran chown -R on /usr/local/bundle. This recursively modified permissions for tens of thousands of system gem files unnecessarily (since they only need to be read at runtime).
3. Sequential processing: Bundler installed gem dependencies sequentially, underutilizing modern multi-core host CPUs.

   This is where I consulted my AI coding assistant.

PROMPT:

Act as an experienced DevOps engineer and suggest the best CPU-centric optimizations for the below Dockerfile. 
I want to keep the changes minimal.

---

The Fix

1. Removed Redundant Installs: Deleted the duplicate gem install command, relying solely on the base image’s Bundler.
2. Optimized CPU Usage: Configured Bundler to use parallel installation jobs matching the host’s CPU cores:

   RUN bundle config set jobs $(nproc) && bundle install
   

3. Scoped Directory Permissions: Restricted the slow recursive chown to /srv/jekyll (which only contains a couple of configuration files during build), keeping the pre-installed gems read-only.
4. Bootstrapped Startup: Updated the Docker Compose startup command to run bundle install dynamically. This immediately writes the correct Gemfile.lock back to the host machine without breaking the runtime environment.

---

The Math: How Much Time We Saved

Since I introduced Docker to this project in this commit in February 2024, I have logged 497 Git commits, including this commit, which contains the time saving Dockerfile updates :

I estimate that :

• Estimated Builds (80% of commits): ~398 builds
• Old Build Time: 96.0 seconds
• New Build Time: 63.5 seconds
• Time Saved Per Build: 32.5 seconds

⏱️ Total Time Saved: 3.6 hours (12,935 seconds) across (398) builds.

My take away, you can take too!

By tuning parallel compiler jobs and narrowing down disk I/O operations, we eliminated hours of developer friction. Sometimes the best DevOps enhancements aren’t new tools, but polishing the ones you already have.

So if you are not that Sr. DevOps mate, ask your Agent to be!

Here is a comparison Before and After (today)

Before (Old Setup)	After (Optimized)
FROM jekyll/jekyll WORKDIR /srv/jekyll COPY Gemfile Gemfile.lock ./ RUN bundle install EXPOSE 4000 CMD ["serve" ..]	FROM ruby:3.2.0 RUN apt-get update && \ apt-get install -y nodejs && \ rm -rf /var/lib/apt/lists/* WORKDIR /srv/jekyll RUN adduser --disabled-password dockeruser COPY Gemfile Gemfile.lock* ./ RUN bundle config set jobs $(nproc) && \ bundle install RUN chown -R dockeruser:dockeruser /srv/jekyll USER dockeruser EXPOSE 4000 CMD ["serve" ..]