Operate the store

After the Deployment chapter, KV ships as a release, runs in containers and forms clusters on Kubernetes. Now it's in production, and things go wrong there that never went wrong on a laptop. This chapter is about finding and fixing them on the running system, with the tools the BEAM ships with.

The concept pages of this track cover those tools one at a time: hot code loading, tracing, memory, mailboxes, schedulers, profiling, crash dumps, and giving AI agents safe, read-only access. The step below puts several of them together on KV.

1. Debug it live. A KV node ignores updates. Reproduce the report with rpc, look at the replica's state, trace one update to find the bug, and hot-patch the fix into the running node without losing its data.

Running these examples

Like the Deployment chapter, each step is a small Mix project whose commands ran when the site was built. Use Download project to get it with the page as its README.