Santekno
Projects/Voucher Seller Services — Deployment Stability & Release Ownership
ByteDance logoPart of ByteDance · Software Engineer, Promotion Platform Team
Backend / APILive

Voucher Seller Services — Deployment Stability & Release Ownership

Owned safe, well-documented production releases for voucher seller services throughout 2024 — feature-flagged changes, prepared hot-config, and continuous monitoring for high uptime.

2024-mvc-deployment-stability · main
# E-Commerce / Release Engineering & Reliability
$ go -X POST /2024-mvc-deployment-stability/v1/sync
→ 201 Created
{ "status": "ok" }
publish 2024-mvc-deployment-stability.synced
Role
Release & Deployment Owner (POC)
Timeline
2024
Team
5
Domain
E-Commerce / Release Engineering & Reliability
Stack
11 tech
Status
Live
01

Overview

Voucher Seller Services — Deployment Stability & Release Ownership was a reliability and release-engineering effort focused on getting frequent backend changes into production safely. I acted as the release/deployment owner (POC) for the voucher seller services, responsible for the end-to-end deployment process: release documentation, feature-flag gating, hot-config preparation, production monitoring, and rollback readiness.

A backend has no screens, so its "UI" is the operational surface: the release documentation, the deployment/rollout flow, and service health dashboards.

Release doc — deployment checklist & feature-flag matrix
Ops view — deployment pipeline & rollout monitoring
Grafana — service uptime & deployment health
03

Problem & Solution

Problem

Voucher seller services changed continuously throughout 2024, and every change had to reach a live, high-availability service without causing disruption. Deployments needed to be well-documented, reversible, and observable — otherwise a single risky rollout could degrade uptime and slow the whole team.

Solution

I owned the release process for each deployment. I authored detailed release documentation describing what shipped and how to verify it, ensured all required changes were feature-flagged, and prepared hot-config configurations before each deployment so behaviour could be adjusted safely. I monitored every production rollout, provided timely status updates to the team, and kept rollback readiness in place. Over the year this delivered 99.99% uptime, 141 pull requests shipped to maintain the services, and 45 of 58 deployments (77%) owned to a successful completion.

04

Workflow

Three actors interact with the service; the diagrams are written in English to stay consistent across both languages.

flowchart diagram

feature-flag + hot-config

deploy

observe

anomaly?

restore

Code Change

Build & Release Doc

Production

Monitoring

Rollback / Flag Off

Release pipeline interaction
sequence diagram
MonitoringProductionFeature Flags / Hot-ConfigBuild PipelineRelease OwnerMonitoringProductionFeature Flags / Hot-ConfigBuild PipelineRelease Ownerprepare release + docsgate changes, stage configdeploy serviceemit metricshealth statusflag off / rollback if needed
Deployment sequence flow
05

Architecture

flowchart diagram

metrics

alerts

Source Repository

Build & Release Pipeline

voucher-seller-service

Feature Flags / Hot-Config

Managed RDS

Prometheus / Grafana

Release Owner

System architecture
Architecture notes

Changes to voucher-seller-service flow through a build-and-release pipeline. Every change is gated by feature flags, with hot-config staged ahead of deployment so rollout can be toggled without a redeploy. The service persists state in a Managed RDS and emits metrics to Prometheus/Grafana, giving the release owner live visibility into deployment health and a clear signal for when to flag off or roll back.

state diagram

feature-flag + hot-config ready

release to production

observe metrics

healthy

anomaly detected

prepared

flagged

deployed

monitored

stable

rolledback

Deployment lifecycle
06

Features

Release Process
  • Detailed release documentation per deployment
  • Pre-deployment readiness checklist
  • Feature-flag gating for every change
Configuration
  • Hot-config prepared prior to each deployment
  • Configuration validated before rollout
Reliability
  • Production deployment monitoring
  • Timely status updates to the team
  • Rollback readiness for fast recovery
07

Tech Stack

language
Go
framework
RPC frameworkHertz
protocol
RPCHTTP
database
Managed RDS (relational database)
release
Feature FlagsHot-Config
infra
Multi-region
observability
PrometheusGrafana
08

Engineering Challenges

01
Shipping many changes to production safely

A high volume of changes had to reach production across the year without destabilising a live service. Fix: a repeatable release process with detailed per-deployment documentation and a readiness checklist, so every rollout was predictable and reviewable.

02
Keeping every change reversible

Any change could regress in production and needed a fast, low-risk way back. Fix: all required changes were feature-flagged and hot-config was prepared before deployment, so behaviour could be toggled or reverted without an emergency redeploy.

03
Sustaining high uptime with minimal disruption

The service had to stay near-continuously available while releasing frequently. Fix: close production monitoring with Prometheus/Grafana, timely team updates, and rollback readiness, sustaining 99.99% uptime across 45 of 58 deployments (77%).

09

Outcomes

99.99%
service uptime
141
pull requests shipped
45/58
deployments owned (77%)
10

Lessons Learned

Reliable delivery comes from process discipline more than heroics: clear release documentation, feature-flagging every change, preparing hot-config in advance, and watching the right dashboards. Making each deployment reversible and observable is what keeps uptime high while still shipping frequently.

11

Related Santekno Tutorials

Interested in working together?

Order, payment, or high-traffic event pipelines — let's talk architecture and trade-offs.