Qovery - Deploy Without Waiting for DevOps

Our DevOps setup was a mess. Every time we needed a staging environment, it would take weeks. Database setup was worse - nobody wanted to touch the RDS configs because the last person who did that broke something for 3 hours.

Tried setting up K8s ourselves for about 6 months. Pretty sure we drove our best engineer insane with all the YAML files - he left for a startup that just uses Vercel for everything. Our monitoring fell over during some big traffic spike and customers were complaining on Twitter before we even knew the site was down. AWS also updated something in EKS and broke our setup for half a day, naturally during a demo to investors.

What This Thing Actually Does

Qovery is basically "Heroku but it runs in your AWS/GCP/Azure account instead of costing $500/month for a tiny app." You push code to Git, it builds your Docker container, spins up the infrastructure, and deploys everything. Your data stays in your cloud account, so you're not locked into anyone's platform.

Why You Might Actually Want This

Deploy Without Tickets: Push to git, get a working deployment. Usually takes like 10-15 minutes. No waiting for the DevOps team to review your load balancer configs or asking nicely for a staging environment.

It Actually Sets Up Kubernetes Right: Setting up production K8s takes months if you want monitoring, logging, security, and auto-scaling that doesn't break at 2am. They've done the work so you don't have to.

Your Cloud Account, Your Data: Unlike Heroku where you're stuck on their expensive infrastructure, everything runs in your AWS/GCP/Azure account. You can export all the configs and leave anytime.

Real Cost Savings: Instead of paying Heroku $25/month per dyno (which adds up fast - we were burning $800/month), you're looking at maybe $200-300/month total AWS costs for the same workload. We went from that $800 Heroku bill to about $280/month on AWS. The math actually works out.

How It Actually Works (No Bullshit)

Connect your GitHub repo, push code. It builds a Docker container, creates a K8s cluster if needed, sets up the load balancer and monitoring. Usually takes around 5-10 minutes depending on how much stuff your app installs.

If something breaks, you get actual error messages instead of "deployment failed, check logs" with no logs to be found. When you need a database, it provisions RDS/Cloud SQL instead of making you figure out persistent volumes and backup strategies.

The catch? You still need to understand Docker basics. If your app needs 16GB of RAM because someone wrote a memory leak, Qovery won't magically fix that. It just makes the infrastructure parts not suck.

Recent Updates Worth Knowing About

They added Karpenter support in early 2025, which actually saves money by not keeping t3.xlarge instances running for apps that need 500MB of RAM. The deployment pipeline got overhauled too - you can finally see what's broken instead of staring at a loading spinner for 20 minutes wondering if npm install hung again.

The deployment queue feature stops that annoying issue where two people try to deploy to the same environment at once and everything breaks.

Real Companies Using This

Talkspace (the therapy app) and some European companies are running production workloads on this. They've got SOC 2 and HIPAA compliance if that matters to you, but honestly the bigger deal is that your data stays in your own cloud account so compliance is mostly your cloud provider's problem.

When This Isn't Right for You

If you have a dedicated DevOps team that actually knows their shit and enjoys debugging pod networking at 3am, skip this. If you're doing ML workloads with GPUs or need custom Kubernetes operators, you're stuck with raw K8s anyway. Also, if you're just hosting a static site, this is probably overkill - Netlify or Vercel are way cheaper.

The Architecture (Without the Marketing Bullshit)

Three main pieces: their control plane (runs on their servers), the deployment engine (open source), and your cloud stuff (runs in your AWS account). Their control plane has gone down a couple times - maybe twice last year for like 30 minutes? - but your apps keep running since they're in your cloud account.

Control Plane: The web UI and API you interact with. Runs on their servers, talks to your cloud account via APIs. Think of it as the remote control for your infrastructure.

Deployment Engine: The actual code that provisions infrastructure and deploys apps. It's open source, written in Rust, and you can run it yourself if you want.

Your Cloud Stuff: Kubernetes clusters, RDS databases, load balancers - all runs in your AWS/GCP/Azure account. You pay the cloud provider directly, Qovery just manages it for you.

What Happens When You Deploy

They upgraded the deployment pipeline in early 2025 because the old one sucked at showing what was actually happening. Now you can see each step instead of staring at a spinner wondering if it crashed:

1. Build your Docker image (couple minutes usually, unless you're one of those teams with 500 npm dependencies, then grab coffee)
2. Provision infrastructure if needed (new K8s cluster takes maybe 8-10 minutes)
3. Deploy your app (rolling update, usually quick)
4. Run health checks and route traffic (or fail with some timeout error if your health endpoint is broken)

If two people try to deploy to the same environment, it queues them instead of failing randomly. Rollbacks actually work now - one click and you're back to the previous version.

The Kubernetes Abstraction (What They Hide and What They Don't)

Look, they don't try to completely hide Kubernetes. You can still get to the underlying configs if you need to. But they set up the boring stuff for you:

• Security: Network policies so your pods can't talk to random shit, RBAC so not everyone is cluster-admin
• Monitoring: Prometheus, Grafana, and log aggregation that actually works out of the box

• Auto-scaling: Horizontal pod autoscaling based on CPU/memory, plus node scaling with Karpenter so you're not paying for idle EC2 instances
• Backups: Automated RDS snapshots, not the sketchy "backup to S3" scripts you were planning to write someday

If you're a K8s expert and want to mess with the Helm charts or Terraform configs, you can. It's not a black box.

How You Actually Use It

Web UI: Pretty typical dashboard - shows your apps, deployment status, logs, metrics. Works fine, not amazing but gets the job done. You can trigger deployments, see resource usage, check logs.

CLI: Pretty useful for automation and debugging. Has a command to drop you into a debug pod when things are broken, which beats the usual cycle of "check logs, guess what's wrong, redeploy, repeat until 3am."

API: Standard REST API if you want to integrate with your existing tooling. Has client libraries for the usual languages.

Terraform: If you're already managing infrastructure as code, they have a provider so you can define Qovery resources in Terraform alongside your other cloud stuff.

Security Stuff That Actually Matters

All traffic is encrypted, certificates are managed automatically so you don't get paged at 3am because Let's Encrypt renewal failed and your site is showing SSL errors. VPC peering works if you need to connect to existing infrastructure.

Role-based access control that doesn't suck - you can give developers deploy access without making them cluster-admin. Integrates with your existing SSO if you're using SAML.

They have SOC 2 and HIPAA compliance, but honestly the bigger deal is that your data stays in your cloud account. Makes compliance audits way easier because you're not explaining why patient data is sitting in some third-party's infrastructure.

What It Actually Costs to Run

The Karpenter integration they added does save real money by scaling nodes based on actual pod requirements instead of keeping oversized instances running 24/7. You can set policies to automatically shut down dev/staging environments outside business hours.

The platform gives you recommendations when you're wasting money. Like "hey, this Node.js app is using 200MB RAM but you allocated 2GB because you panicked during the last memory spike."

So you're looking at $29/month per developer plus whatever you spend on AWS. For a small team that might be like $200-500/month in AWS costs, but I've seen teams hit $1200/month because they left every staging environment running with production-sized databases. The Qovery subscription is usually the cheap part - it's the database instances that'll eat your budget. Still probably cheaper than hiring a full-time DevOps person, assuming you can find one who actually knows what they're doing.