Abhibhaw Asthana

Work / BharatPe · Internal developer platform

BharatPe: 14 days to 17 minutes

Engineers waited 14 days for a small DevOps team to run database queries and server changes. I built a self-service portal with access rules and audit logs.

was 14 days now 17 minaverage wait for a request
20,000+requests served
User
Engineers who needed a database query run, data access, or a server change.
Problem
A small DevOps team did every request by hand. Engineers waited 14 days on average, and nobody could audit who got access to what.
Decision
A self-service portal. Engineers request what they need, the right approver accepts it, and the portal runs it with access rules and an audit log.
Not built
Open access for everyone, or a bigger ticket queue. The portal runs approved actions under rules, so the DevOps team reviews rules, not every request.
Result
20,000+ requests through the portal. The wait fell from 14 days to 17 minutes.

How it works

Before
  1. Engineer
  2. raises a ticket
  3. removed: DevOps queue14 days on average
  4. done by hand
  5. Database or server
After
  1. Engineer
  2. requests, with a reason
  3. Approvermanager or data owner
  4. accepts
  5. Portalruns it under access rules
  6. logs it
  7. Audit trailin Jira
The queue left the path. Approval, access rules and an audit trail replaced it.

Engineers choose from supported actions: database queries, Kafka, GitHub, Redash and API gateway requests. Risky actions, like changing a live MySQL table, need a second sign-off from DevOps.

What I left out

Direct access for everyone was fast but unsafe. A bigger queue was safe but slow. The portal runs approved actions under rules instead.

What happened

  • 20,000+ requests went through the portal.
  • The average wait fell from 14 days to 17 minutes.
  • Every grant has a requester, an approver and a time.

One part is written up in detail: automating Redash permissions.