Posts

Disclaimer: This post was written with the help of Claude Sonnet 4 You know that feeling when you’ve automated everything perfectly, and then reality hits you with a curveball? That’s exactly what happened during our recent migration from media.tifi.tv to media.telex.im. What started as a straightforward file migration turned into an unexpected deep dive into Elasticsearch operations. Here’s how it all unfolded. The Perfect Plan (That Wasn’t So Perfect) The mission seemed straightforward: migrate all media files from our old domain to the new one. We had thousands of files scattered across our PostgreSQL database—user avatars, uploaded documents, logos—all pointing to https://media.tifi.tv/telexbucket/. The new infrastructure was ready at https://media.telex.im/ with separate buckets for staging and production environments. ...

We all have our moments when we want to learn something cool. It struck me this evening and I decided to pick up Rust again. It feels more natural this time due to my more extended knowledge of systems engineering. Now std::program::exit(1) doesn’t seem like Rust magic anymore but a standard way programs should exit gracefully. So I decided to implement Two Pointers in Rust. Boy did I forget what two pointers was. I had to ask GPT to describe two pointers in 10 words. It said something along the lines of, find two numbers in a list that sum to a target number. Now, I chose to design my program to receive all the inputs as program args. The first thing I learnt to do was collect a program’s inputs. Luckily Rust has a handy way in the standard library: std::env::args().collect(). This creates a vector of all passed arguments. Now the next thing to learn was how to parse the collected arguments. Rust prefers pattern matching to for-loops. So I picked up this piece of code from ChatGPT: ...

hi When you place one end of a metal spoon in hot water, heat flows from the end submerged in water to the end that is not. The temperature at each point in that spoon when viewed collectively is a scalar field. If a temperature-camera were to capture this spoon, it would look like this. ...

Disclaimer: this post is AI generated. Promtail is a powerful log agent designed to ship logs from local files to a Loki instance. If you’re setting up a logging pipeline with Grafana Loki, proper Promtail configuration is crucial. This guide walks you through the essential steps to configure Promtail effectively. Understanding the Configuration Process Configuring Promtail involves three key steps: Writing a proper configuration file Ensuring proper file access permissions Restarting the service to apply changes Let’s dive into each step in detail. ...

If you were asked to package HTML (and other static assets) in a Docker container, how would you achieve it? Sure, you can spin up a Node.js or Python application and create a web server that serves the required assets, but that would be a roundabout method of solving a simple problem. Of course, Nginx is a viable option, but which will you choose? $ docker image ls REPOSITORY TAG IMAGE ID CREATED SIZE nginx latest 53a18edff809 7 weeks ago 192MB nginx stable-alpine-slim 95ebc828d689 7 weeks ago 11.8MB ...

A robot needs to know how to: Localize itself Plan its paths Control itself Robots are built to do three kinds of jobs: Dirty Dangerous Dull Robots generally rely on three technologies: Sensors - for perceiving their environment Actuators - for locomotion and movement Algorithms - to guide their decision making Robots may face these three challenges: Efficiently managing energy - energy efficiency Adapting to new environments - environment adaptability Interacting with humans and other living organisms - human interaction Three general types of robots: ...

Training machine learning models is one thing, using them is another. When models are trained for real-time inference, you need a runtime that can deliver results with milliseconds accuracy. If you trained your model on a PyTorch pre-trained model, your inference time will not be optimal if you simply use the torch state dictionary during inference. Converting your Pytorch model into a faster runtime like ONNX is a faster alternative. In this blog post, you will learn how to convert a Pytorch state-dictionary model into ONNX format for faster inference. ...

When running virtual machines in Azure, you sometimes need to add additional IP addresses to your VM’s network interface. While Azure typically manages IP assignments through its network interface configuration, there are situations where you might need to configure IP addresses directly on the VM level. Introduction Azure VMs come with a primary IP address assigned through DHCP. This IP is managed by Azure and configured automatically when your VM boots up. However, there are scenarios where you might need to add secondary IP addresses: ...

Disclaimer, this article was AI generated Overview This guide provides detailed instructions for implementing a secure authentication system using NGINX with JSON Web Tokens (JWT). You’ll learn how to install all necessary dependencies, compile NGINX with the nginx-auth-jwt module, and deploy two complementary Go microservices: one for generating and issuing JWTs and another that demonstrates how NGINX validates these tokens in a real-world scenario. By the end of this tutorial, you’ll have a robust authentication system that can be integrated into your existing microservices architecture. ...

Disclaimer: This post was written with the help of Claude Sonnet 4 PostgreSQL, the perfect database that has been adapted to do everything sometimes shouldn’t be used to do everything. PostgreSQL shines for ACID compliance, for relational joins, and its virtual extensibility. But what if you have millions of data that should be searched through in seconds? What if you have situations where data replication across nodes is not an afterthought but part of the fundamental system? What if you want to horizontally distribute resources across multiple machines? In such cases, you might want to explore something without the shackles of PostgreSQL. Something in the league of MongoDB or Elasticsearch. ...