Nine Days to a Kafka TopicPipeline · hard · ~13 minThe platform survey's headline number: median 9 days from "I need a Kafka topic" to a usable topic. The path: a ticket, an infra-team review, a partition-count negotiation, a manually-written Terraform PR, and credentials pasted over a Slack DM.Multiply across the estate: 14 data systems (Kafka, the warehouse, an OLAP store, search, object stores…), each with its own team, queue, and tooling. Engineering leadership has funded a provisioning control plane and you own the design.An architect on the team proposes the quick version: "ship a shared library — every infra team imports it, writes to a shared metadata DB, done in a quarter." A company you all read about tried exactly that. You're deciding the architecture.Evidence

The provisioning tax

The reference case's shared-library postmortem

Requirements

Checkpoints

Why will the shared-library proposal recreate the reference case's four failures?

In the gateway-plus-providers design, which responsibilities belong to the CENTRAL gateway? (Pick the correct set.)

The reference case cut provider onboarding from 12 developer-months to 2. Your estate has 14 systems. At the improved rate, roughly how many developer-months to onboard the full estate?

Write the design's governance spine: how the catalog stays drift-free, how credentials stop flowing through Slack, which request classes auto-approve versus gate, and the contract rule that keeps 14 provider teams from eroding the plane.