Day 3 - May 4, 2026: From Automation to Observability

Day 3 marked a shift from building systems to making them measurable.

The work centered on resolving last-mile deployment friction, restoring a healthy deployment pipeline, and moving the Cloudflare automation work toward an observable BI-driven platform. The system had already started to automate and repair infrastructure drift. The next step was proving what it was doing, surfacing where it was working, and making the results visible outside the CI logs.

Goal / Intent

The goal was to transition the Cloudflare Infrastructure as Code workflow from automated and self-healing to measurable and observable.

That meant pushing infrastructure audit results into Google Sheets, shaping the data for Looker Studio, and beginning the first operational dashboard components for tracking Cloudflare fleet health.

Platform & Deployment Fixes

Before the observability work could be trusted, the deployment path needed to be clean again.

Scan and CI issues were creating workflow reliability friction, so the first step was to troubleshoot the failing checks and confirm which failures were quality-gate problems versus deployment connectivity problems.

The key deployment issue was that Cloudflare Pages was still tied to an old GitHub repository that had been deleted. That stale association left the site in an unclear state: the project existed in Cloudflare, but the deployment source was no longer a valid repository target.

The fix was to re-establish a clean GitHub connection in Cloudflare and reconnect Pages to the current 100daydash.blog repository. Once the repo ownership path was clear again, the deployment pipeline was restored.

This reinforced three platform requirements:

• Deployment traceability matters. A deployed site needs an obvious source repo, branch, and build path.
• Repository ownership must stay unambiguous. Deleted or renamed repositories should not remain hidden dependencies in deployment systems.
• CI/CD observability is part of reliability. A passing workflow is useful, but knowing what is connected, when it ran, and what it deployed is what makes the platform operable.

Work Completed

CI/CD & Automation

The automation controls were tightened around explicit remediation.

Manual remediation now requires dual gating:

• run_remediation=Y
• FIX_DETECTED_GAPS=Y

That keeps write-enabled repair paths intentional while allowing read-only audit workflows to run safely on a schedule.

The audit workflow now supports weekly scheduled checks on Sunday UTC, while Google Sheets synchronization can be triggered on demand with workflow_dispatch. The split is deliberate:

• read-only audit workflows identify current state and drift
• write-enabled remediation workflows require explicit operator intent

Data Pipeline & Architecture

The Google Sheets integration moved to a dual-table design:

• recent stores the latest snapshot and is overwritten on sync
• history stores append-only records for time series analysis

This data shape is intentionally simple. The recent sheet supports current operational views without requiring Looker Studio to calculate the latest record per domain. The history sheet preserves the time series needed for trend analysis.

Pre-shaping the data upstream avoids complex “latest timestamp” logic in BI and lets the dashboard support both real-time and historical analytics from the same pipeline.

BI Integration

Looker Studio was connected to both Google Sheets datasets:

• recent powers scorecards and the operational domain view
• history powers trends and time series analysis

The first calculated fields were added:

• Compliance Level, mapping domain state to Pass or Fail
• Security Configuration, combining SSL and bot protection signals

The first dashboard components are intentionally basic but useful:

• Fleet Health Scorecard
• domain-level status table

Some early BI friction also surfaced. ISO 8601 timestamps needed careful parsing, and UTC timestamps had to be treated consistently against local viewing context. Those issues are manageable, but they confirmed that timestamp modeling should be solved upstream where possible instead of hidden inside chart configuration.

Documentation

Documentation was updated to match the current system behavior.

The README now clarifies workflow triggers, documents the Sheets integration, and explains the remediation safety controls. That improves onboarding for this project and makes the Cloudflare workflow easier to reuse across related projects without rediscovering the same operational rules.

Key Insights

Pre-shaped data is better than complex BI logic.

Separating snapshot data from historical data simplifies the dashboard model. The operational view can ask, “What is true now?” while the historical view can ask, “How is the fleet changing over time?”

Looker Studio’s limitations are useful feedback. They reinforce the need to model data upstream in a way that is easy to query, chart, and explain.

The system is now evolving through a clear platform arc:

Infrastructure as Code -> Self-Healing -> Observable Data Platform

Definition of Done

Day 3 was complete when:

• Cloudflare and GitHub were reconnected through a stable Pages deployment path
• CI/CD workflows were functioning for manual and scheduled execution
• Google Sheets sync was working correctly
• Looker Studio was connected to both recent and history
• the initial scorecard and domain table were working accurately
• documentation reflected the current workflow behavior

Next Steps

• Add a time series compliance visualization for the “Hero’s Journey” view.
• Track fleet and domain growth over time.
• Add compliance distribution charts.
• Improve dashboard layout, labeling, and visual hierarchy.
• Introduce compliance recovery rate across pre-remediation and post-remediation states.
• Start aligning the dashboard narrative with portfolio storytelling.

Reflection

Day 3 was about moving from building systems to proving they work.

Automation creates leverage, but visibility turns that leverage into something communicable. The Cloudflare workflow is still early, and the Looker Studio dashboard is not yet mature, but the shape is now clear: infrastructure state can be audited, repaired, synced, measured, and explained.