louis gueye 💉💉💉

@Pure_Grit_ @dhh Yep and for me, nothing is simpler than linux. Really works for my usage. Had to use both mac and windows at work. Nothing compares to linux when it comes to being truly productive.

@dhh Tech loyalty? It's all about what makes life simpler, really.

I remember a time when techies would say "I have to use Windows at work, but at home I use OSX!!". Now it's "I have to use a Mac at work, but at home I use Omarchy!!" I hear from people who love computers. Funny how history rhymes!

@ulrich Pas mieux ici. Les enfants étaient ravis

Marre de ce beau temps qui m’oblige à me vautrer dans les BBQ !

@davidfowl @digitalocean has a very good developer experience

What is your favorite place to deploy personal projects outside of the big 3 cloud platforms?

@kozlovski So the said metadata now live somewhere in the broker disk space? Just curious... In any case it's really a nice operational simplification: no need for an additional zookeeper in testcontainers tests

After 14 years of service, ZooKeeper will be officially gone from Apache Kafka starting… next week. 👀 By now, you should know about KRaft. It is Kafka's implementation of Raft - its protocol for achieving distributed consensus. ✨ This consensus is used to: • determine the order of metadata changes • elect a single Kafka Controller which is responsible for updating said metadata and reacting to changes in it (e.g. electing new partition leaders when a broker dies) In a few bullet points, the way KRaft works is the following: 1. you have a few controller brokers who use Raft to elect a leader amidst themselves 2. the leader becomes the active Kafka Controller. All metadata changes are persisted in an internal Kafka topic, which the active controller leads. 3. every Kafka broker consumes this topic, and that is how they learn about the latest metadata 💡 🤨 What was the problem with ZooKeeper? 1. scalability - loading a lot of data from ZK was expensive and slow. Large Kafka clusters would also have a lot of writes to ZK, and ZK simply wasn’t made for such throughput. 🔥 2. availability - the slower you load data, the greater the unavailability window when a Kafka controller fails. KRaft is a 15x (!) improvement in some hard-failure cases. 🤯 3. maintainability - different project means different experts. There are not many people in the world that are both ZooKeeper and Kafka experts. Who do you find to triage and solve ZooKeeper bugs? 🔧 How do you build your Kafka team around this? 4. simplicity - it’s simpler to run and secure one system (Kafka) rather than two (Kafka & ZK) 🛡 🏰 Four Major KRaft Pillars You Should Remember 🧠 Separation between Controllers and Brokers Unlike before, only a few brokers are now eligible to become the active controller. As part of deploying Kafka, you configure precisely which brokers are eligible to become controllers. They can be co-located with a normal broker as part of the same JVM, or they can be their separate processes (on their own separate nodes). This allows you to isolate, scale and upgrade the controllers separately. 🧠 Records all the way down & Incremental Updates Every metadata change is captured in a plain old Kafka topic (metadata topic) with plain old Kafka records that represent the delta change from the last state. The leader of this topic can only be one of the controller brokers. It is chosen via the Raft consensus protocol. Whoever is the leader is the Active Controller. 🧠 Push → Pull Model The topic is also actively consumed by ALL brokers. This is how they maintain the latest state of metadata. Brokers now pull (fetch) data from the Controller, instead of the Controller sending requests to them. 🧠 Hot Standby Since everything is modeled in a regular Kafka topic and each record is an incremental state delta, an inactive controller already has a vast majority of the latest state loaded on its local disk. In the previous model, because everything was stored in ZK, a controller that just restarted had to fetch all teh data over the wire. Now, the controller simply reads it from its disk! Only tiny bit is fetched over the wire - the changes which occurred while the controller was down. 👌 😰 Is It Stable? You tell me: • KRaft has been marked production ready since November 2022 (Kafka 3.3). • No major blockers have been reported recently • Confluent announced completing the largest ZooKeeper->KRaft migration in the world, moving its entire Confluent Cloud fleet (thousands) onto KRaft "without downtime" 🎬 Fin These changes significantly improve Kafka's efficiency. For example, the previous upper bound on the total number of partitions a cluster could have was around 200,000. With KRaft, this is now lifted. 🥳 More importantly, KRaft paves the way for many more improvements down the road! It is a very elegant solution to distributed consensus that embodies the principles behind event streaming much more natively. Just as Kafka should. 👍 If you're still running Kafka with ZooKeeper, the time to migrate is NOW.

Je vais vous faire gagner de l'argent et du temps : ça demande bcp de travail. Faut en bouffer des exos et des exos pour avoir le niveau 😉

@Shirleyyych Totalement crevé quand ils avaient 0-10 ans. L'entrée au collège change tout. Donc désolé mais t'en as encore pour quelques années. Après le collège par contre c'est nous qui quémandons un peu d'attention: on est totalement ignorés

@Shirleyyych Ne m'en parle pas: du nom de la société en passant par les noms+prénoms de personnes ainsi que leur fonction. Certains ne comprennent pas le concept de discrétion et d'anonymisation ... Et c'est pas des juniors...

VLC automatic subtitles generation and translation based on local and open source AI models running on your machine working offline, and supporting numerous languages! Demo can be found on our #CES2025 booth in Eureka Park.

@Shirleyyych Merci pour ce post. C'est toujours très intéressant de pouvoir mettre des mots sur un comportement observé inconsciemment... Ça permet d'en parler de façon plus crédible!

🇪🇪 La ville de Tallin va planter un arbre pour chaque bébé né dans la capitale ! 3500 enfants devraient naitre dans la ville en 2025, soit autant d'arbres pour verdir ses rues ! Le projet vise à améliorer le cadre de vie et atténuer les îlots de chaleurs. 👶 (err) 📸 Pixabay

This man bought a car damaged by mudslide and it took him a week to clean it!

@Shirleyyych Très bons rappels. On le fait peut-être spontanément mais, là c'est très bien abordé. Je recommanderai ce short à beaucoup de candidats !

@SumitM_X You need a config server which is the ground truth of the configuration of your workloads (n micro services). Then a config server client deployed on each runtime interested in a config change. That way, whenever the server updates the config, client workloads can reload it

You need to dynamically adjust the configuration of your microservices (e.g., changing database connection pools or timeout settings) without redeploying the services. How would you achieve this?

[𝗘́𝗩𝗘́𝗡𝗧] Prepare for trouble... Make it double! The Clever Team is back for a second day at Forum PHP! 👋 Come and meet Ismaël Cissé, @allema_s, @PerreuxFlorent, @davlgd and Carine Guillemet on our booth.

@SumitM_X You can use MinIO to store the file. Then use the urn in the payload to reference the file. Also stream while reading from and writing to MinIO to keep the memory at an acceptable level.

Your microservice needs to transfer large amounts of data (e.g., files, images) between services. How do you design the communication to avoid performance bottlenecks and manage large payloads efficiently?

@sivalabs @venkat_s Same goes for filing open source issues: - never forget that those guys contribute on their own non paid time - provide context, even better, take the time to provide a reproducible executable - if you can't provide either, politely wait until they take the time to fix the issue

Knowing how to ask for help is a must-have skill. If you are going to send an email or Twitter DM asking for help, do yourself a favor by reading this amazing blog post written by @venkat_s to learn how to ask for help in a professional manner. Always be professional when you write an email blog.agiledeveloper.com/2017/10/always…

@Shirleyyych Perso je m'étais intéressé à Dahmer, j'ai regardé 2 épisodes et j'ai arrêté, c'était trop pour moi. Du coup suis super refroidi sur Monsters. Je pense pas avoir le courage

@sivalabs @sdkman_ That, plus I can manage multiple project with different jdk version without any issue. So comfortable...

@lgueye @sdkman_ Same here. Once I started using @sdkman_, I never had to download and install JDK manually, and I was not worried about configuring JAVA_HOME env variables, etc.

IntelliJ IDEA Tip 💡 @sdkman_ is an amazing tool to install different versions of JDK, Maven, Gradle, JBang, etc. You can use ".sdkmanrc" file to specify which versions of SDKs you want to use and run "sdk env" to activate those versions in the current shell. java=21-tem maven=3.9.9 Even better, you can set "sdkman_auto_env=true" in "~/.sdkman/etc/config" file so that when you navigate to the directory having ".sdkmanrc" file, those versions will be activated automatically. @intellijidea automatically detects ".sdkmacrc" file and configures the project-level JDK automatically.

@22feno Oui, clairement, dans le foot je valorise plus les passeurs qui sont de l'ordre du génie en QI football que les buteurs. Sans minimiser le travail des buteurs, c'est tellement plus simple de conclure quand on a reçu un ballon de Beckham, Modric, Inesta, Pogba ou Kroos...

On devrait changer un truc dans le foot genre si la passe est de ce niveau là normalement c’est le passeur qui va célébrer hein