Guides

Investigating production incidents

Methods that work with whatever tools your team uses, and how ForgeOps does the same checks for you.

Did a deployment cause this incident?

How to tell whether a deploy really caused a production incident: four checks on timing, release, changed code and everything else that changed, and why the latest deploy is so often the wrong suspect.

Read the guide →

Correlating errors with deploys and database queries

How to line up errors, deploys and database queries during a production incident: one timeline, following a failing request to its slowest query, telling a slow query from an exhausted connection pool, and handling missing data.

Read the guide →

How to investigate a production incident

A step-by-step method for investigating a production incident: measure the impact, stop the damage, list what changed, test hypotheses against evidence, follow one failing request, and write the timeline as you go.

Read the guide →

See an investigation step by step

Real incidents replayed minute by minute, with the evidence changing the answer as it arrives.

Watch an incident replay