Lessons learned with FHIR Paging and Plain/Façade FHIR/Load Balanced-Systems
summary
HAPI FHIR’s sample paging keeps search results in memory which breaks behind load balancers and forces eager loading; implemented a working alternative that caches search criteria (not patient results), uses a custom BundleProvider plus a paging provider to persist a serializable arguments object to Redis, marks IOC references transient and resurrects beans by name, and sets paging TTL (example 60 minutes) and max defaults. Demo shows stored UID (5747) and Redis blob, and the page-2 fetch returns rows on demand — it works in a load-balanced, pay-as-you-go way.
AI-written summary · about
chapters (10)
- 0:00Intro
- 2:27Basic Anatomy (next)
- 3:10FHIR Server as a Façade Observations
- 5:55Hapi Fhir Observations (Documentation)
- 6:15Hapi Fhir Observations (Paging Providers)
- 9:21Load Balance Issues (Alternate Idea)
- 11:29Load Balance Issues (Custom Bundle Provider)
- 14:57Load Balance Issues (Custom IPaging Provider)
- 15:38Time/Space/Continuum Issues
- 16:38Time for Demo? (Time allowing)
