Preparing for Disaster: Handling the Election and MGP at NRK.no
summary
NRK.no describes concrete preparations and failures: they run on Kubernetes with auto-scaling, use caching and an event mode for DDOS protection, and rely on Firebase for live interactions. Known events (1.14M viewers for Musikkmša/Maskorama, 17M election pageviews) get load-testing and scaled replicas, but unexpected peaks and a 100% CPU hit on Firebase plus an under-scaled Ingress controllers during Melodi Grand Prix cause voting delays and an “emoji storm” outage; the voting later works and the result remains valid.
AI-written summary · about
chapters (26)
- 0:00NRK.no introduction
- 1:09NRK.no ecosystem
- 3:14News website complexity
- 4:18High-traffic event handling
- 7:08Election night role
- 9:16Breaking news surges
- 11:46Unplanned disaster prep
- 14:37Scaling unpredictable events
- 17:27Pandemic as disaster
- 19:34War traffic and DDOS
- 21:43Pre-event prep services
- 22:48Load testing system knowledge
- 24:12Monitoring during events
- 24:55Mistakes and postmortems
- 27:01Failure stories lessons
- 27:222019 election load test fail
- 28:052021 election missing service
- 29:30Monsoon TV mapping failure
- 30:55TV production communication gaps
- 31:17Chemistry finale crash
- 32:41Winter Olympics DDOS false positives
- 34:05MGP 2012 voting meltdown
- 40:44MGP 2012 aftermath lessons
- 43:12Kubernetes bottlenecks
- 45:42Backups in critical moments
- 46:03Q&A
related talks
11mServing election results to an entire nation
52mNot all attacks are malicious: DDoS from outside and within
05mIgnite - What I Learned From a Dress an On-Call Nightmare
20mWhat I Learned From A Dress, an On-Call Nightmare
10mSurviving the 2008 Elections
36mX-Factor Voting Platform Under the Hood
06mWhat I Learned From A Dress, an On-Call Nightmare
24mRollout on Reception: Progressive Delivery for GenAI Apps Using Real...