Stacked Diffs on @GitHub will start rolling out to early design partners in an alpha next month. In the meantime, here's video of our progress so far:
(h/t for @georgebrock + team for their awesome work)
You do a live DB migration.
1. Create a new table with the new structure.
Write to the old table, them to the new table. Consistent write to old table. You read from the old table.
2. Open reading from the new table, with fallback to old, for selected users. Check system works.
3. Open reading from new table, fallback to old to all users. Check system works.
4. Point of no return - shift writing to new table first, then to old table. Consistent write to new table.
5. Migrate all missing rows to new table on your own schedule. No rush.
6. Drop old table.
🎄 Jour 16 du Calendrier de l'Avent des 🐻
Xan : un outil en ligne de commande Open Source, écrit en Rust au médialab de SciencesPo 🇫🇷, conçu pour manipuler de très gros fichiers CSV directement depuis le terminal.
Un outil qui nous a été chaudement recommandé par d'admirables devs ❤️
Our Kubernetes cluster was using 400% more CPU than needed.
The symptoms:
- Nodes constantly at 80% CPU
- Frequent pod evictions
- Slow application response times
- Auto-scaling triggering constantly
The investigation:
- Profiled applications with pprof
- Found inefficient JSON parsing in hot path
- CPU-intensive logging at debug level
- Unnecessary data transformations
The fixes:
- Switched to faster JSON library
- Reduced log verbosity in production
- Cached computed values
- Optimized data structures
Result: CPU usage dropped to 20%, response times improved 3x, infrastructure costs down 60%.
Application efficiency matters more than infrastructure scaling.
Our team moved from LastPass to self-hosted Vaultwarden.
The breaking point:
- LastPass had another breach
- $6.5 per user per month
- 15-person team = $1170 annually
- Limited control over data
- Compliance team wasn't happy
Why Vaultwarden:
- Open source Bitwarden server
- Written in Rust, lightweight
- Single Docker container
- Uses 10MB of RAM
- All Bitwarden features for free
The setup:
- Deployed on t3.micro in private VPC
- Behind Application Load Balancer
- Let's Encrypt for SSL
- PostgreSQL for data storage
- Automated backups to S3 every 6 hours
- Took 3 hours total including testing
Cost breakdown:
- EC2: $8/month
- RDS: $15/month
- Total: $23/month
- Savings: $1147 annually
The surprise benefits:
- Full audit logs of who accessed what
- No user limits for free
- Emergency access configuration
- Custom password policies
- Data stays in our infrastructure
18 months later:
- Zero downtime
- Team actually uses it
- Compliance audit passed easily
Checkov : un scanner Open Source d'infrastructure as code pour détecter les vulnérabilités et les mauvaises configurations.
"Policy-as-code for everyone"
👉 le projet : https://t.co/Od0hgh31Dr
👉 En savoir plus : https://t.co/16n7cMZMn0
Stripe 💗 intense traffic spikes. Last week, @Zevent held the largest charity stream in France—€10m raised!—and smoothly handled traffic peaking at a sustained 3,000 requests per second with 99.9999% uptime. (If you try 3,000 rps on most payments providers, they will fall over.)
🐻 Les 10 commandements de l’éco-conception d’infrastructures chez Bearstech
Du choix du DC à l'optimisation des archis, découvrez nos 10 commandements pour allier performance, sobriété énergétique et durabilité dans vos infrastructures.
https://t.co/9tn1sL4g4u
I just got my third security audit back. All clear.
Most backend devs think HTTPS + JWT = secure. That's how you get owned.
Here are the security measures that actually work ↓
@GergelyOrosz@Pragmatic_Eng Didn't know about the goalie!
We have a similar role at Shotgun: the Chief Bug Officer (CBO).
An engineer fixes bugs and answer tech requests for a week.
So there is always someone fixing stuff.
And we got to $200M GMV/yr without an on-call rotation thanks to this CBO role.
Atlassian switched from JSON to Protobuf for one of their services, resulting in:
- 75% less memcached CPU usage
- 80% smaller data
- 33x faster deserialization
- a 55% smaller memcached cluster
this guy is getting paid $35,000 to set up an internal “ChatGPT” for a law firm.
> locally hosted Llama for LLM
> N8N to connect it all.
we’re living in the AI gold rush.