Nine teams were taking turns on one staging environment

说明与用途

多支团队共享单一预发布环境导致频繁故障和归因问题。采用每个拉取请求的隔离环境减少了错误，加快了调试，并将成本降至每月300英镑。

问题

A single shared staging environment caused frequent failures and attribution problems when multiple teams deployed changes, leading to costly manual debugging and slow release cycles.

结果

Reduced staging failures, eliminated attribution ambiguity, cut manual debugging time, and lowered cost to £300/month while maintaining shared resources via locks.

工作流

• Identify need for isolated environments
• Define environment in code (namespace, DB snapshot, stubs, hostname)
• Automate creation on PR creation
• Run soak tests for ~2 h
• Destroy environment on merge or after 12 h
• Use pipeline locks for shared resources

证据

We spent most of a morning investigating a staging failure in checkout…

来源

A namespace, a database restored from a sanitised nightly snapshot…

来源