A 20‑Minute Adventure: Issues and Patches During OK Streaming Upgrade
summary
Describes how Odnoklassniki's (OK) streaming system is architected and recounts the concrete problems, patches, and failures encountered during a framework upgrade after 10+ years of evolution. Covers the streaming technologies used at OK and the practical fixes applied while updating that streaming framework.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 4:39Samza upgrade challenges
- 6:05Kafka authorization issues
- 8:34Kafka offset errors
- 10:20UI Application Master problem
- 11:24First production rollout
- 12:47Transactional checkpointing issue
- 14:16Data loss from partitioning
- 16:21Task and partition problem
- 19:14State loss from retention
- 21:42Full state loss on Friday 13
- 25:19Summary
- 26:24Q&A
related talks
24mPeople Make the Difference or Video Streaming at Odnoklassniki
48mA 4K Streaming Platform for a Million Concurrent Viewers
39mFailure Testing
33mThe Challenges of Testing Video Playback at Scale
24mPushing Your Streaming Platform to the Limit
38m10 years of OSMC: Why does my monitoring still look the same?
35mA Service with Cameras - a Performance Minefield
44mRelease After a Year of Development