I'm curious about the less common but highly effective tweaks you've made to your infrastructure. For example, I recently optimized our Dockerfiles to use specific version numbers instead of just 'latest'. This change greatly improved the stability of our deployments. What are some other unique changes that have really made a difference for you?
5 Answers
We analyzed our costs and noticed that intra-AZ traffic was a big expense, especially since we use Kafka a lot. After weighing the risks, I suggested moving everything to a single AZ. It's a gamble, but so far it’s led to lower costs without many negative impacts. Fingers crossed!
We run Java apps on Kubernetes and introduced multilayer images. By layering our base image, JVM, dependency libraries, and the actual application, we created smaller build images. This speeds up the build process and saves space on the image repository, making it more efficient overall.
I've been wondering if there's a significant benefit to separating the JVM image and the application. Care to elaborate?
We switched to using snapshot and release versioning for our application. This means instead of rebuilding artifacts right before deployment, we now have clean, reproducible artifacts that work consistently across dev to prod. This has made our deployments a lot smoother!
That's interesting! Could you share what exactly you snapshot? I'd love to understand better.
This approach used to be common before everyone jumped on GitHub actions and artifact recreation for deploys.
One impactful change we made was automating the on/off schedules for our non-prod cloud resources. We built a tool that integrates with Terraform and ArgoCD to power down infrastructure after hours and turn it back on when needed. This simple change has cut our cloud spend by 25-60%!
I revamped our Confluence documentation, even though no one specifically asked for it. I knew it would benefit my team, and it did! Documentation can become so outdated, so I've also made it a rule that if you need to follow certain documents, they should always be kept up to date.
Sounds like a smart move! Keeping documentation fresh is so crucial.
Wow, I wish I had a setup like that! I've heard about proper documentation but haven't really seen it.
Have you looked into a cell-based architecture? That might balance cost with resilience!