Running the Infra at FOSDEM
summary
FOSDEM runs ~10,000 people per day and ~80–90k devices on volunteer-operated infrastructure, delivering live streams and Wi‑Fi without proprietary services. The stack uses IPv6-only with DNS64, monitoring with Prometheus, Loki and Grafana forwarded to an offsite backup, and video handled by a local rendering farm replacing Blackmagic encoders with commodity laptops. Operational wins: switched from BIND to CoreDNS for large-scale DNS performance, Ansible rebuilds in ~30 minutes, and a 10 Gbps BGP uplink that required removing an MTU misconfiguration to stabilize.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:46FOSDEM as volunteer-run conference
- 3:31Core infrastructure setup
- 5:20Video streaming workflow
- 8:13Volunteer coordination system
- 10:00Network challenges at university
- 11:48IPv6-only network with DNS64
- 13:17Conference build-up timeline
- 15:47Video setup and hardware management
- 17:32Post-COVID infrastructure recovery
- 19:1710Gbps link and BGP troubleshooting
- 21:01Monitoring and backup systems
- 22:25MTU issues and resolution
- 23:31BGP session migration test
- 24:13Future bandwidth plans
- 24:34Public dashboards and documentation
- 25:37Q&A session
related talks
01h 48mOpenInfra Summit 2022 Berlin: Keynote Day 2
38mRunning a FOSS associative ISP
20mConnecting, Securing and Optimising Enterprise Networks with Free Software
19mDeveloping, Deploying, Operating In-House Network Switches at a Massive Scale
17mState of Open Source Networking and Edge - Day 2
01h 33mDay 3 Part 2 Sessions
25mLet’s make local and accountable tech!
01h 02mHow companies use LOKI to run their business