4.5. Reflection Checkpoint

Key Takeaways

• Operating compute resources is about visibility and incremental control over already-running fleets — connecting, viewing, snapshotting, and scaling — a distinct skill set from Phase 3's deployment decisions.
• Data operations split into querying (day-to-day interaction), backup/restore (protection against loss, with restores targeting a chosen instance rather than always overwriting the source), and cost/fleet visibility tools like Database Center.
• Networking operations are incremental adjustments — resizing subnets, reserving static IPs, adding routes — layered on top of the topology designed in Phase 3, not a redesign.
• Monitoring (is something wrong now) and Logging (what exactly happened) are complementary; Cloud Trace, Profiler, Query Insights, and index advisor each target one specific diagnostic symptom, not observability in general.
• Telemetry collection is never automatic — Ops Agent, custom metrics, and audit log configuration all require deliberate setup before they produce any data.

Connecting Forward

Phase 5 closes the loop with the domain that touches every other one: configuring the access and security controls — IAM policies and service accounts — that govern who and what can perform every action described in Phases 2 through 4.

Self-Check Questions

• Without re-reading 4.1.1, can you explain why a scenario about standardizing new VM creation points to images rather than snapshots?
• If a teammate assumes Cloud NAT lets external users reach a private instance the way a load balancer would, what's the specific correction from 4.3.2?