Cloud-native Postgres observability: from client apps to underlying cloud resources

How do you maintain reliable observability over PostgreSQL databases in cloud environments where systems span hundreds of services, dozens of databases, and infrastructure that shifts dynamically under your feet? At Prague PostgreSQL Developer Day, Peter Zaitsev, Coroot Co-Founder, explored practical strategies for keeping visibility intact in exactly those conditions.

Peter covered how eBPF can be used to generate comprehensive service maps and instrument Postgres calls from any client application to specific databases, giving teams a clear picture of how their data layer fits into the broader system. He also addressed automatic discovery mechanisms for topologies within High Availability Postgres clusters managed by different Kubernetes Operators, removing the manual effort that complex HA setups typically require.

Beyond the database itself, Peter examined the importance of monitoring the underlying cloud resources, including virtual machines, storage volumes, and network components, that Postgres depends on for stable, predictable performance.