Reading logs without panic

When a pipeline turns red, the instinct is to change something. The better first move is to read the log from the bottom.

Find the first error, not the last

The final line is usually a symptom: “job failed”. The cause is the first line that looks wrong. Scroll up until the output stops being normal, then read forward from there.

Notice how long it took

A step that fails after one second never really started. Look at checkout, shell setup and the very first command. A step that fails after twenty seconds got past setup and ran real work. The duration alone narrows the search.

Ask what is different

Something that worked yesterday and fails today has changed. Was it a commit, a secret, a DNS record, a tool version? A short list of recent changes beats guessing.

Reproduce locally

If your checks are scripts, run the same script on your machine. A failure you can reproduce is half fixed. A failure that only happens in CI points at the environment: a different shell, missing tool or missing permission.

Write the fix down

The second time you hit the same error should take a minute. Keep a troubleshooting note with the error text you saw and what fixed it.