Site Reliability Engineer

About Tinybird: At Tinybird, we help developers and data teams take flight by unlocking the power of real-time data to quickly build data pipelines and innovative data products. With Tinybird, you can effortlessly ingest multiple data sources at scale, query and shape it using the 100% pure SQL you already know and love, and publish results as low-latency, high-concurrency APIs for your applications to chirp about. Developers can create fast APIs, faster—what used to take hours and days now only takes minutes! Tinybird is the essential tool that data engineers and software developers have been waiting for enabling you to drive innovation with ease.

What you will be doing: We are looking for someone to help us scale and to keep our software and infrastructure reliable and elastic as we scale. Someone who knows how to make hardware and software play together, and that participate as part of the oncall team, to understand not only our product, but also the issues our clients face.

‍We run our stack in Linux. We try to keep things simple. Technologies we use:

OpenResty: SSL termination and load balancing

Varnish: load balancing and, sometimes, caching

Redis: metadata store

Python: most of our backend uses Python except some small bits that rely on C++ for hot paths

ClickHouse: our main data store

Zookeeper: for ClickHouse replicas coordination

We use Grafana, Loki and Mimir for monitoring and alerting

Terraform: Cloud provisioning (virtual machines, networks, Kubernetes clusters)

Ansible: Deploys and software and config provisioning

Our number of machines is still manageable, but the number keeps growing as we keep adding customers. This is not about managing infrastructure but about making sure that our software uses the hardware resources wisely and flexibly. This means you will not only have to worry about automating machines, but about helping the product team to design and develop the architecture of the system as a whole. That will require you to work with our backend code and to understand how ClickHouse works.

Some challenges and things we want to improve:

High-availability and elasticity: as we keep adding customers, we need to architect our system to be more efficient and flexible

Observability: from specific resource usage to a bird's eye view of the whole platform. This requires good knowledge of storage, networking, and computing

Disaster recovery: improving our tooling to manage and discover problems, but also improving our on-call procedures

As a specific challenge: when our customers grow, we need to upgrade their accounts. Now, we do it manually—not in the traditional sense of manual because we have tools that automate much of the process, but we need to take care of that one customer at a time: deciding what machines we need to spin-up, how much compute capacity we will provision, etc. Ideally, our architecture should allow our customers to upgrade themselves and assign more resources to them dynamically and seamlessly in the most dynamic, safe and transparent way possible.