DevOps Rescue: CI Red, Docker Stuck, Main Not Live

Live is the SHA on the host, not the PR

origin/dev is not production. A green check on a topic branch is not the container. We match git, the registry tag, and curl.

Usual jams

Runner offline. concurrency cancel-in-progress killed the deploy. Lint on a hosted runner that has no minutes. Secrets missing in Actions. Healthcheck on the wrong port.

What we do

Read the failed job, not the Slack rumor. Fix the gate or the runner. Ship the known-good image. Prove HTTP 200 on the public host.

After the fire

You can keep us for fleet ops — patches, SSL, pins — as an Oniyore retainer.

Questions

Do you rewrite the whole pipeline?

Not for a stuck release. We get this SHA live, then leave a residual ticket for the architecture.

Self-hosted runners?

We start or queue; we do not silently fall back a fail-closed lint onto a random box.

More notes

• Website Down Today: What We Fix in the First Hour
• SSL Certificate Expired: Browsers Block, Calls Stop
• Contact Form Not Sending: The Quiet Lead Leak
• DNS Cutover Same Day Without a Mystery Outage