Reliability is our number one issue at Stripe
In this video, the Stripe team discusses their approach to building a platform with Temporal that is different from how many other teams do it.
Key points from the video:#
- Distributed ownership: Temporal allows horizontal scalability and isolation, so Stripe can distribute ownership of worker services across their large engineering org.
- Fat client strategy: To lower the cost of service ownership, Stripe asks teams to set up a worker service to run their workflows. They’ve wrapped the Temporal SDK and hid it from their customers to avoid needing a bunch of institutional knowledge of Temporal.
- Safety measures: Stripe has implemented a number of safety measures to protect their users from classic blunders, including forcing developers to choose retry policies for activities and statically detecting calls to functions that could cause replays to fail.
- Versioning: Stripe forces people to version their fields when adding them to a grpc interface to prevent errors caused by changing the history replay.
Overall, Stripe has taken a unique approach to building a platform with Temporal that is focused on reliability, safety, and ease of use for their developers.
