Infrastructure

A controlled runtime layer

I start by inventorying services, dependencies, domain traffic, component purposes, and their relationships. Linux users, access permissions, container boundaries, configuration files, and secret locations can then be brought into a coherent structure.

Reverse proxies receive explicit routing rules, automatic certificate renewal, and only the public ports they require. I make changes in small steps after recording the system baseline and preparing a recovery plan.

Recovery, observation, and patching

A backup is useful only when its scope, storage location, and restoration path are understood. I plan backups for data, configuration, and essential keys, then use recurring restore tests to expose gaps in the procedure.

Monitoring covers service availability, resource usage and limits, certificate expiry, and meaningful application signals. Alerts go to a channel that will be noticed, while log review distinguishes isolated noise from a recurring fault.

Security and system patches are grouped by risk and compatibility. A steady cadence provides room to test changes instead of leaving old components without an explicit decision.

Managing the Cloudflare edge

At Cloudflare, I maintain DNS records, caching rules, WAF filters, and traffic routing. Edge settings remain aligned with the origin so protection does not hide faults and stale private content does not remain cached.

Workers can handle a small part of routing, header, or static content logic close to the visitor. I use them for bounded purposes, observe execution failures, and avoid duplicating behaviour that is simpler and more dependable in the existing system.

Incident habits and durable ownership

When an alert arrives, I confirm the impact, contain damage, and record the timing of actions. Once service is stable, I examine the root cause, related signals, and preventive changes that could reduce recurrence.

A runbook explains access, diagnosis, restart, restoration, and responsibility handover. Architecture notes and the reasoning behind significant choices reduce dependence on one person's memory.

Periodic reviews cover resource usage, abandoned services, storage growth, and the purpose of subscriptions. Cost can then be reduced without quietly removing recovery or visibility.

Care routine

• Review availability, resource, and certificate alerts.
• Apply operating system and service patches on an agreed cadence.
• Inspect backup results and perform restoration exercises.
• Check authentication, application, and reverse proxy logs.
• Record Cloudflare DNS, WAF, cache, and Workers changes.
• Update runbooks, architecture notes, and the access inventory.
• Assess capacity, load patterns, and recurring infrastructure spend.

When to reach out

• Alerts arrive too late or have no consistent owner.
• Backups exist but no one has proved they can be restored.
• Certificates, containers, or system packages rely on ad hoc updates.
• Incidents end with a temporary patch and no cause analysis.
• Cloudflare rules have accumulated without documentation or intent.
• Infrastructure spending is rising faster than actual usage.