Deployment Health Dashboard

Atlassian has hundreds of microservices. At the time, there was no single "plane of glass" to see the health of deployments across the organization. Engineering managers had to construct mental models by tab-switching between Jira, Bitbucket Pipelines, Jenkins, and Splunk logs. When a bad deployment happened, the "Mean Time To Detect" (MTTD) was high because nobody looked at the logs until a customer complained.

I pitched and built the "Deployment Health Dashboard", an internal tool designed to aggregate build and deployment signals into a single, real-time timeline.

I built the backend using Node.js and Apollo GraphQL. The core challenge was data ingestion. - **Ingestion Service**: I wrote a set of webhooks and pollers to normalize data from GitHub Actions, Bitbucket, and Jenkins into a standard "Deployment Event" schema. - **Storage**: We stored these events in PostgreSQL. Since we needed to show historical trends ("Are our deployments getting slower over time?"), I designed the schema with time-series queries in mind. - **Caching**: To ensure the dashboard loaded under 200ms, I implemented a Redis caching layer for the heavy aggregation queries (e.g., "Success rate per team over the last 90 days").

The UI was built with React and used Apollo Client for state management. I focused heavily on "Information Radiators", designing the UI to be readable from across the room on a TV screen. We used color-coded status indicators and sparkline charts to show trends at a glance.

The tool was adopted by over 15 engineering teams within a month. It became the default screen in many standups. We saw a tangible reduction in incident resolution time because teams could correlate a spike in errors directly to a specific deployment on the timeline.