Spoorthy is a BEAM language and functional programming enthusiast with 15 years of industry experience, much of it in BEAM languages. Spoorthy has experimented with the BEAM across many domains — Telecom, Finance, Medical, Web, Messaging, Crypto, and more — and has seen its industrial application at major players such as Ericsson, Apple, and Mastercard. Spoorthy has also contributed to the Erlang open source library OTP and to the community.
Talk is about Designing Supervisor Trees That Actually Recover and also When the Supervisor Can’t Save which is failure ways that OTP doesn’t catch much. Designing supervisor trees tells Beyond the basics of one_for_one, talk digs into sophisticated supervision strategies for production systems, covering rest_for_one, dynamic supervisors, circuit breakers built from GenServers, and the difference between a system that restarts and a system that genuinely recovers with preserved state. Having said on the design can also reflect upon what Supervision trees doesnt handle —like ETS table bloat, message queue floods, and cascading timeout storms and couple others. This talk catalogs the production failure modes that OTP’s built-in mechanisms don’t address, and presents the monitoring, alerting, and architectural patterns that do. Would be demonstrating examples from my own industry experiences where i have personally faced them
Target Audience: