Prisma delivers a platform that offers developers all the primitives to develop robust and scalable backend applications, including data storage, reactivity, long-running workflows, and APIs. Building upon the success of its popular open-source ORM, Prisma is dedicated to delivering an exceptional developer experience for those working with databases in applications deployed on serverless and the edge. The Prisma Data Platform comprises a suite of products designed to streamline the development of Serverless applications.
The Challenge: Gather troubleshooting info quickly
The complex microservice applications that developers build with Prisma need both a debugger and an observability platform. Teams must troubleshoot identified problems, but they also need to discover unknown unknowns before those spawn more trouble.
When Prisma customers contact support, Prisma’s support team needs to identify the underlying symptoms and situation quickly, and determine if Prisma is meeting internal, as well as external expectations. They then need to provide Prisma’s engineers with as much specific context as possible to debug the problem and provide a solution. The longer that end-to-end process takes, the less faith customers will have in Prisma — even if it turns out Prisma wasn’t the problem.
Prisma support staff often need to spin up quick data sketchpads — temporary queries, graphs or dashboards to collect troubleshooting info. Tools like Grafana provide stable dashboards for the long term, but it often took support staff too long to create something quick to help a customer troubleshoot.
The Solution: Axiom regex support and APL
Axiom proved much quicker than other tools for verifying performance metrics observed by a customer or for building dashboards on the fly, says Marco Ieni, a Prisma software engineer. This removes the latency in getting initial troubleshooting info that can frustrate a customer waiting for an answer — and saves Prisma staff their own time.
How does Axiom do this? It lets them use regular expressions with its Axiom Processing Language (APL) to extract, say, a user ID and latency from logs much more quickly than using a more formalized approach. Ieni says the regex approach is much faster than other tools to quickly explore datasets. Axiom’s query performance lets them skip the need for a pipeline.
Prisma support staff, operating with read-only access, create queries they can share with engineering. This removes the maddening back-and-forth messages between customers, support and engineering as they try to collect debugging data or test solutions. Instead of trying to explain things in text, they can share a query and its results. It cuts both latency and frustration for all involved.
Finally, Axiom’s pricing is based on ingest terabytes rather than the number of logs or events. This keeps the Prisma team from having to worry they’ll run up a surprise bill by doing their jobs. They can collect the data they need as they need it, to help customers succeed and help Prisma to iterate a better product.