CDL
Deep Dive
Cloud Monitoring, Cloud Logging, Cloud Trace, Cloud Profiler, Error Reporting, and the SLI/SLO/SLA framework for managing service reliability.
Google Cloud Observability (formerly Cloud Operations Suite and, before that, Stackdriver) is Google Cloud's integrated observability platform. It covers the three pillars of observability: metrics, logs, and traces.
| Service | What it does | When to use |
|---|---|---|
| Cloud Monitoring | Collects metrics from GCP resources, AWS, and your own apps. Creates dashboards, alerts, and uptime checks. | Monitor VM CPU, API latency, endpoint availability. Alerts when metrics exceed thresholds. |
| Cloud Logging | Centralizes logs from all GCP services and your own apps. Log search, analysis, and export. | Diagnose production errors, access auditing, compliance. Export to BigQuery for analysis. |
| Cloud Trace | Distributed tracing for APIs and microservices. Shows the time of each component in a request. | Identify latency bottlenecks in microservices apps or apps with multiple API calls. |
| Cloud Profiler | Performance analysis of production code. Identifies functions that consume the most CPU or memory. | Optimize production code without affecting performance. Detect memory leaks. |
| Error Reporting | Automatically groups and counts application errors. Real-time notification of new errors. | Quickly detect new types of errors after a deployment. Prioritize which bugs affect the most users. |
Metrics and alerts
• GCP metrics: CPU, memory, disk, network, API latency — automatic with no configuration
• Custom metrics: your own business metrics (orders/minute, revenue, active users)
• Alerting policies: notify by email, PagerDuty, Slack, or Pub/Sub when a metric exceeds a threshold
• Uptime checks: verify URL availability from multiple global locations
Cloud Logging: retention and export
• Retention: 30 days by default for most logs. Configurable up to 3650 days.
• Log Router: export logs to BigQuery (analysis), Cloud Storage (archive), Pub/Sub (real time)
• Log-based metrics: create metrics from patterns in logs (e.g., counting 500 errors)
• Audit Logs: Admin Activity (who changed what), Data Access (who read what), System Events
This framework, originated by Google's SRE (Site Reliability Engineering) team, defines how to measure and manage the reliability of production services.
SLI — Service Level Indicator
The actual metric you measure. The quantifiable way to evaluate the performance of the service.
Example
"99.2% of HTTP requests returned a response in less than 200ms over the last 30 days."
SLO — Service Level Objective
The internal performance target. The team's internal agreement about what level of SLI is acceptable.
Example
"99.5% of requests must respond in less than 200ms." — If the SLI is below the SLO, there is a problem to fix.
SLA — Service Level Agreement
A contractual commitment to the customer. If breached, there are consequences (credits, penalties).
Example
"Google guarantees 99.9% monthly availability for GKE. If not met, service credits."
Key relationship: SLO stricter than the SLA
The internal SLO should be stricter than the SLA committed to customers. If the SLA is 99.9%, the internal SLO should be 99.95%. The buffer is the "error budget" — margin for deployments, maintenance, and incidents without violating the SLA.
| Concept | Definition | In GCP |
|---|---|---|
| High availability (HA) | A system designed to minimize downtime. Redundancy across multiple zones or regions. | Multi-zone Managed Instance Groups + Cloud Load Balancing + Cloud SQL with automatic failover. |
| Fault tolerance | The system keeps working even if some components fail, without manual intervention. | GKE auto-healing pods. Cloud Spanner multi-region with 99.999% SLA. |
| Disaster recovery (DR) | The plan and capacity to restore services after a major failure (an entire region down, data corruption). | Backups in multi-region Cloud Storage. Cloud SQL automated backups. Persistent Disk snapshots. |
| RTO (Recovery Time Objective) | The maximum acceptable time to restore service after a disaster. | Defines how long you can wait. Active-active multi-region = RTO of seconds. Backup and restore = hours. |
| RPO (Recovery Point Objective) | How much data loss is acceptable. "How far back can we go?" | Cloud SQL with PITR (Point-In-Time Recovery). Frequent backups = lower RPO. |
| Autoscaling | Automatically adjusting resources based on load. Horizontal (more instances) or vertical (more CPU/RAM). | MIG Autoscaler, GKE HPA, Cloud Run scale-to-zero, BigQuery automatic scaling. |
DevOps
A culture and set of practices that unite Development (Dev) and Operations (Ops) teams to accelerate software delivery with higher quality and reliability.
SRE (Site Reliability Engineering)
A practice created by Google: software engineers applying engineering principles to operations. The goal is to maximize the reliability and availability of systems at scale.
Google Cloud offers different support plans based on business needs. Support is part of the operational experience and can be decisive in minimizing the impact of incidents.
| Plan | What it includes | Who it's for |
|---|---|---|
| Basic | Documentation, community forums, access to the Google Cloud Status Dashboard. | Exploration projects, development, early-stage startups. |
| Standard | Unlimited technical support by email during business hours. P2 response in 4 hours. | Non-critical production workloads that can tolerate short interruptions. |
| Enhanced | 24/7 support for critical incidents. P1 response in 1 hour. Technical Account Manager (TAM) available. | Critical production apps. Businesses where downtime directly impacts revenue. |
| Premium | P1 response in 15 minutes. Dedicated TAM. Proactive architecture reviews. Event management. | Large enterprises with mission-critical workloads. Very strict business SLAs. |
Did you get this topic?
Test what you just learned
The team of an e-commerce app wants to be automatically notified when the HTTP 500 error rate exceeds 1% for 5 minutes. Which Google Cloud service should they use?
Put what you learned to the test with original CDL exam questions.
Start CDL mock examUnlimited mock exams with the Pro plan
Log in to track your progress.