Production Reliability Check.
A read-only review of your release path and alerting.

Most teams we talk to suspect something is off in their release path or their monitoring, but cannot say what, or how much it matters. The Check answers that before anyone commits to a longer engagement. It is fixed in scope and price, it is read-only, and it ends with a plan you can carry out with us, with someone else, or on your own.

The release path

• Whether production runs the build that passed your checks: moving tags, digests and stale images on nodes.
• Whether a failed deploy turns the pipeline red.
• How a rollback would happen today, and whether your database migrations allow it.
• Which secrets CI holds, and whether it needs them.
• Which checks gate a production deploy, and whether they still run on every change.

More on this side in deploy safety for Docker on VMs.

The alerting

• Alerts whose queries return no data, so they can never fire.
• Alerts that fire but reach no one.
• Noise: how many alerts closed themselves over a recent period, and which hosts raise them.
• Whether anything watches the monitoring itself.
• What you pay for monitoring, set against what the team actually uses.

More on this side in observability for Docker on VMs.

• A written report. Each finding with the evidence behind it, ranked by risk.
• A 90-day plan. What to fix first, in what order, and what each step involves.
• The fee back if we do the work. If you hire us for the work the plan recommends, the fee for the Check is credited against that first engagement.

1. Scope call. Thirty minutes to agree what is in scope. The price is fixed and agreed before any work starts.
2. Read-only access. To the monitoring account and the CI configuration, plus an export of recent alerts or incidents. We change nothing in production.
3. Review. About two weeks from access to report.
4. Report and plan. We take you through the findings and the 90-day plan, and answer your team's questions.

Every item on the list is something we have found in production: a Swarm deploy that brought back the previous build with CI green; alert conditions watching metrics that did not exist; a real alarm that reached nobody; 95% of a manufacturer's monitoring tickets closing themselves. The studies below show what we found and the work that followed.

It is not a penetration test or a compliance audit, and it does not change anything in production. We do not promise a number of findings: if your release path and alerting are sound, the report says so and shows why.