Hung integration test blocked hotfix deploy for over three hours at checkout
A one-line fix for a checkout bug that rejected a card type was merged in August but took nearly four hours to reach production. The delay was caused by a shared GitHub Actions runner labelled 'deploy' that was simultaneously running a slow integration test stuck in a blocking read. The test had connected to a message broker undergoing maintenance, which accepted the connection but never responded, and with no read timeout, no per-test limit, and no job timeout set, it inherited the platform's six-hour default. Engineers discovered the issue at 5:30 PM by checking what the runner was actually doing, and the hotfix deployed within nine minutes of cancelling the stuck test. The team has since separated deploy and test runners, enforced explicit job timeouts via linting, added per-test timeout limits, and set up on-call alerts for any deploy job queued longer than five minutes.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in