ฝึก GCP Cloud DevOps Engineer

Cloud Build compiles source and builds container images. Cloud Run runs containers, Cloud Storage stores objects, and Cloud DNS resolves names.

Cloud Deploy automates delivery pipelines and progressive rollouts. Cloud Functions is serverless compute, Cloud SQL is a database, and Cloud CDN caches content.

Cloud Source Repositories provides private Git repositories. Storage stores objects, Logging stores logs, and Monitoring observes metrics.

Artifact Registry stores container images and language packages. Functions is compute, SQL is a database, and DNS resolves names.

Cloud Build triggers start builds when code changes are pushed or when schedules fire. Tables, load balancers, and DNS records are different.

A Dockerfile contains instructions for building a container image. Shell scripts, schemas, and network policies serve other purposes.

Infrastructure as code manages resources through declarative or imperative configuration. Documentation, spreadsheets, and manual clicks are not IaC.

Deployment Manager deploys Google Cloud resources from declarative templates. Monitoring, Logging, and Trace are observability services.

Config Sync applies Git-based configuration to GKE clusters. Building images, monitoring, and DNS are different functions.

Cloud Build builds images, Cloud Deploy manages pipelines, Artifact Registry stores packages, and Source Repositories hosts Git.

SRE applies software engineering to run reliable systems. Marketing, design, and payroll are not operational engineering practices.

An error budget is the allowed failure budget derived from an SLO. It is not a financial budget, backup, or quota.

An SLO defines reliability targets such as 99.9 percent availability. Project names, firewall rules, and storage classes are different.

An SLA is an agreement with expected service levels and penalties. Images, DNS zones, and indexes are not agreements.

An SLI is a measured metric such as availability or latency. Regions, machine types, and billing accounts are not reliability metrics.

Canary releases test a new version with a small traffic share. Deleting versions, creating projects, or disabling monitoring are not canary practices.

Rolling deployments update instances in stages to reduce downtime. Replacing everything, creating accounts, or stopping traffic are not rolling updates.

Blue/green runs old and new environments and shifts traffic when the new version is ready. One environment or deleting environments is not blue/green.

Load tests measure behavior under traffic or usage. Backups, key rotation, and image builds are not load tests.

SLO defines a target, error budget is allowed failure, canary uses small traffic, and blue/green switches environments.

Cloud Monitoring collects metrics, dashboards, and alerts. Logging stores logs, SQL is a database, and Storage stores objects.

Cloud Logging ingests and queries logs. Cloud Run and Functions are compute, and CDN caches content.

Cloud Trace analyzes latency and request traces. Profiler analyzes CPU and memory, DNS resolves names, and Armor protects web apps.

Cloud Profiler captures continuous CPU and memory profiles. Trace analyzes latency, Build builds images, and Deploy manages rollouts.

Cloud Error Reporting aggregates and surfaces application errors. Monitoring tracks metrics, DNS resolves names, and SQL is a database.

Uptime checks test endpoints from multiple locations. Backups, firewall rules, and quotas are unrelated.

Alerting policies evaluate conditions and notify teams. DNS records, buckets, and images do not alert.

Log-based metrics count or measure patterns in log entries. CPU counters, billing reports, and DNS queries are different.

Dashboards display metrics and charts for operational visibility. Logs, load balancers, and images are not dashboards.

Monitoring collects metrics, Logging stores logs, Trace analyzes latency, and Error Reporting groups errors.

Secret Manager stores and controls access to secrets. Storage stores objects, SQL is a database, and CDN caches content.

Cloud IAM grants roles and permissions. DNS resolves names, Trace analyzes latency, and Profiler analyzes CPU.

Service accounts let applications authenticate to Google Cloud APIs. Passwords, firewall rules, and DNS records are different.

Artifact Analysis scans container images for vulnerabilities. DNS resolves names, Logging stores logs, and Functions runs code.

Cloud Armor provides WAF and DDoS protection. CDN caches content, DNS resolves names, and Trace analyzes latency.

Binary Authorization enforces that images are signed and approved before deployment. Backups, billing, and load tests are unrelated.

Firewall rules allow or deny traffic based on source, destination, ports, and protocols. Buckets, machine types, and tables are not firewall rules.

Cloud Audit Logs records admin, data access, and policy events. Cloud Logging stores general logs, Trace analyzes latency, and Profiler analyzes CPU.

Cloud KMS creates and controls encryption keys. IAM manages permissions, DNS resolves names, and Armor protects web apps.

Secret Manager stores secrets, IAM controls permissions, Binary Authorization enforces signed images, and KMS manages keys.

Deployment strategies define how changes reach users, such as rolling or canary. Schemas, DNS records, and firewalls are unrelated.

Progressive delivery uses canaries, feature flags, and gradual rollout to reduce risk. Deleting changes, creating users, or disabling logging are not progressive delivery.

Feature flags control feature availability without redeployment. DNS records, storage classes, and machine images are different.

Rollback restores a previous version after a failed deployment. Backups, scaling, and key rotation are separate operations.

A release candidate is a build that may become the production release. Tables, firewall rules, and buckets are not release candidates.

Pipelines automate stages such as build, test, and deploy. Routes, DNS zones, and backups are not pipelines.

Promotion moves approved artifacts between environments. Deletion, project creation, and password changes are unrelated.

GitOps manages deployments from Git repositories. Documentation, spreadsheets, and email are not GitOps.

Canary analysis checks whether a new version meets criteria before expanding traffic. Deleting, creating projects, or disabling alerts are not analysis.

Feature flags toggle features, rollbacks revert versions, GitOps uses Git, and promotions move artifacts.

Cloud Billing tracks spend, budgets, and exports. Trace analyzes latency, Profiler analyzes CPU, and DNS resolves names.

Recommender gives recommendations for cost, performance, and security. Logging stores logs, Functions runs code, and CDN caches content.

Budget alerts notify teams when costs exceed limits. DNS records, firewall rules, and storage classes are unrelated.

Recommender suggests rightsizing VMs based on utilization. Armor protects web apps, Trace analyzes latency, and Build creates images.

Idle resources cost money without delivering value. Firewall rules, DNS zones, and images are not idle resources by definition.

Committed use discounts lower cost in exchange for committed usage. Pay-as-you-go has no commitment discount, and spot VMs are interruptible.

Quotas limit resource usage such as API calls or instances. Backups, DNS records, and load tests are not quotas.

Cloud Monitoring provides utilization metrics used for cost decisions. DNS, Deploy, and Source Repositories serve other purposes.

Cost optimization balances efficiency and performance. Deleting everything, ignoring budgets, or disabling monitoring are not optimization.

Recommender gives recommendations, budget alerts notify, quotas limit usage, and committed use discounts reduce cost.