Team Ownership - Who Owns Production
summary
Claim: giving teams end-to-end ownership works only with clear owners, tooling and coordination. A Node shim plus gradual 2% rollout unexpectedly doubles backend load—one request fans out to 15–20 calls, ~25x traffic—so tracing with Zipkin surfaces the fault and enables fixes. Teams adopt centralized bug triage rotations, embedded SREs in product squads, cross-discipline guilds and a web-infra glue team to manage shared libraries and reduce firefighting. Hiring remains the right scale-up when tooling and processes can’t keep up.
AI-written summary · about
chapters (13)
- 0:00Introduction to talk
- 1:03What is Reddit
- 3:31Reddit's early growth
- 4:55Team structure evolution
- 6:42Production ownership challenges
- 8:07Solutions for ownership issues
- 10:12Team composition and guilds
- 12:56Node.js shim experiment
- 16:08Debugging the shim problem
- 17:32Consequences of unclear ownership
- 18:54Scaling infrastructure teams
- 19:15Adding embedded SREs
- 20:20Key takeaways and summary
related talks
16mScaling Teams with Ownership
29mWho Owns What: Ownership Culture while Bitzscaling
23mExtreme Ownership for Global Teams
14mCentralising the Right Things
01h 10mAgile Cross Team Collaboration How Tos
01h 02mTools for simplified large scale systems
23mManaging diverse engineers for high-impact deliveries
51mTech-enabled Network of Teams