Cloud & infrastructure

The cloud provider looks after the hardware, the network and, depending on the services you choose, more than that. A managed database can leave backups and some updates to the provider, but you choose how long backups are kept, and if you choose nothing the provider’s default applies, which can be just a few days. We check what the service actually covers and that backups can be restored. What stays with you is the configuration, the permissions, what should happen when something breaks and who should do it.

Why do deployments feel scary?

A deployment feels scary when nobody quite knows what separates the test environment from production. The differences live in the head of whoever set the environments up, or in a setting someone changed by hand and forgot. Every release becomes an event, and nobody holds events for no reason, so releases get fewer and bigger. Environments described in code show how they are meant to look and make the differences deliberate. What is actually running is another matter, so before a release we compare the environments with the code.

What we set up

• Environments defined as code, where the differences between test and production are deliberate and written down, and an environment can be rebuilt from the code.
• An automated build and deployment pipeline, where the previous version can be brought back without a new release. Database changes and side effects in other systems do not roll back by themselves, so they are planned separately.
• Monitoring and alerts with a named recipient. An alert nobody receives is just a log.
• Backups tested by actually restoring them, with the restore time measured.
• Secrets and permissions with an owner, one place where they are managed and a way to rotate them.
• Cost tracking per environment and service, with an alert when something starts costing more than usual.

Cloud or your own hardware?

It depends on the load, how even it is, and what you need to be able to tell authorities and customers about where the data sits. Most of the environments we look after are in the cloud, though your own hardware does come up. Whichever you choose, we want to answer yes to the same question: can the environment be rebuilt without one particular person being reachable? The ongoing work after that is application management.