Google Cloud Professional Cloud Architect Managing Implementation and Reliability Operations Syllabus

How much persistent storage does Cloud Shell provide and what is preserved?

Cloud Shell provides 5 GB of free persistent disk mounted at your $HOME directory; only files in $HOME survive across sessions.

What is a gcloud configuration and why use multiple named configurations?

A gcloud configuration is a named set of properties (account, project, region/zone). Multiple configurations let you quickly switch contexts between different projects or accounts using 'gcloud config configurations activate'.

In automation, what is the benefit of using '--format=json' or '--format=value()' with gcloud commands?

It produces machine-parsable output so scripts can reliably extract field values, rather than fragile parsing of human-readable text.

What is the difference between imperative scripting (gcloud) and declarative infrastructure (Terraform/Deployment Manager) for automation?

Imperative scripts describe the step-by-step commands to reach a state; declarative tools describe the desired end state and the tool reconciles current to desired, making them idempotent and easier to track drift.

What is event-driven architecture and one core benefit?

A design where services communicate by producing and consuming events asynchronously rather than direct synchronous calls. Core benefit: loose coupling—producers and consumers scale and fail independently.

What is the difference between Pub/Sub and Eventarc in Google Cloud event-driven integration?

Pub/Sub is a general-purpose async messaging service (topics/subscriptions). Eventarc is a managed eventing layer that routes events from Google Cloud sources (and beyond) to destinations like Cloud Run using a standardized CloudEvents format.

What is the difference between Pub/Sub push and pull subscription delivery?

In pull, the subscriber requests/pulls messages and acks them. In push, Pub/Sub proactively sends (HTTP POSTs) messages to a subscriber endpoint and expects a success response as the ack.

What delivery guarantee does Pub/Sub provide by default, and what must consumers handle?

Pub/Sub guarantees at-least-once delivery, so consumers must be idempotent to safely handle possible duplicate messages.

What is the difference between a metric, a time series, and a dashboard in Cloud Monitoring?

A metric is a measurable property type (e.g., CPU utilization); a time series is a stream of timestamped values of that metric for a specific resource; a dashboard is a visual collection of charts displaying those time series.

What are the three kinds of metrics in Cloud Monitoring (by value behavior)?

Gauge (value at a point in time), Delta (change over an interval), and Cumulative (a continually increasing total since a start time).

What is a custom metric in Cloud Monitoring and when would you create one?

A user-defined metric you write from your application or agent to track app-specific signals (e.g., queue depth, business KPIs) not provided by built-in Google Cloud metrics.

What is an alerting policy in Cloud Monitoring?

A configuration that defines conditions on metrics (thresholds, durations) which, when met, trigger notifications via channels (email, SMS, PagerDuty, Pub/Sub) to signal a problem.

What is Cloud Logging and what are the two default log buckets?

Cloud Logging is Google Cloud's centralized log management service. Its two default buckets are _Required (retains certain admin/audit logs ~400 days, non-configurable) and _Default (all other logs, default 30-day retention).

What is a log sink in Cloud Logging and what is it used for?

A sink routes (exports) matching log entries to a destination such as Cloud Storage, BigQuery, Pub/Sub, or another log bucket, based on an inclusion/exclusion filter.

What are the four types of Cloud Audit Logs?

Admin Activity, Data Access, System Event, and Policy Denied logs.

Which Cloud Audit Log type is enabled by default and cannot be disabled?

Admin Activity audit logs (and System Event logs) are always written and cannot be disabled; Data Access logs are disabled by default (except BigQuery).

What is the difference between Cloud Trace and Cloud Profiler?

Cloud Trace is a distributed tracing system that shows latency across requests as they flow through services; Cloud Profiler is a continuous, low-overhead CPU and memory profiler that identifies code-level resource consumption.

What does Cloud Trace measure and how does it help diagnose issues?

It collects latency data (spans) as requests propagate across distributed services, helping pinpoint which service or call is the bottleneck causing slow response times.

What is Error Reporting in Google Cloud?

A service that automatically aggregates, groups, and counts errors/exceptions from your applications' logs, alerting you to new or frequent errors with stack traces.

What is an SLI, and how does it relate to an SLO?

An SLI (Service Level Indicator) is a quantitative measure of service behavior (e.g., % of successful requests). An SLO (Service Level Objective) is the target value or range for that SLI (e.g., 99.9% success).

What is the difference between an SLO and an SLA?

An SLO is an internal target for an SLI. An SLA (Service Level Agreement) is an external contract with customers that includes consequences (penalties/credits) if the agreed objectives are not met; SLAs are typically looser than SLOs.

What is an error budget and how is it calculated from an SLO?

The error budget is the allowed amount of unreliability: 100% minus the SLO target. For a 99.9% SLO, the error budget is 0.1% of requests/time that may fail before the objective is breached.

What is an error budget policy and what typically happens when the budget is exhausted?

It is an agreed policy governing actions based on error budget consumption. When the budget is exhausted, teams typically freeze new feature releases and prioritize reliability work until the budget recovers.

Roughly how much downtime per month does a 99.9% (three nines) availability SLO permit?

About 43.2 minutes per 30-day month (approximately 8.76 hours per year).

What is the difference between horizontal and vertical scaling (scale out vs. scale up)?

Horizontal scaling (scale out) adds more instances/machines to share load; vertical scaling (scale up) increases the resources (CPU/RAM) of a single instance.

What is a Managed Instance Group (MIG) and how does it enable autoscaling?

A MIG is a group of identical VM instances created from a template. Its autoscaler automatically adds or removes instances based on metrics like CPU utilization, load balancing capacity, or custom metrics.

What are the main types of Google Cloud Load Balancers and a key distinction between them?

Global external HTTP(S), SSL Proxy, and TCP Proxy load balancers (Layer 7 / proxy, global), versus regional external/internal Network (TCP/UDP) pass-through load balancers (Layer 4). HTTP(S) LBs route by content/URL; Network LBs forward by IP protocol/port.

What is the difference between a global and a regional load balancer in Google Cloud?

A global load balancer (e.g., external HTTP(S)) uses a single anycast IP to route users to the nearest healthy backend across regions; a regional load balancer serves backends within one region only.

What is a health check in the context of load balancing and autoscaling?

A periodic probe that determines whether a backend instance is healthy and able to serve traffic; unhealthy instances are removed from rotation (and, with autohealing, recreated).

What is a quota in Google Cloud and what are the two main quota types?

A quota limits resource consumption to protect against runaway usage. Two types: rate quotas (requests per time period, e.g., API calls/minute) and allocation quotas (number of resources, e.g., VMs per region).

What is throttling (rate limiting) and how should a client respond to a 429 'Too Many Requests' / 'RESOURCE_EXHAUSTED' error?

Throttling rejects requests exceeding a rate limit to protect a service. Clients should back off and retry later using exponential backoff with jitter rather than retrying immediately.

What is the difference between a cache hit and a cache miss, and how is hit ratio defined?

A cache hit is when requested data is found in the cache; a miss is when it must be fetched from the origin. Hit ratio = cache hits / (cache hits + cache misses).

Compare cache-aside (lazy loading) vs. write-through caching strategies.

Cache-aside: app loads data into cache only on a miss (lazy), risking stale data but caching only what's used. Write-through: data is written to cache and datastore simultaneously on every write, keeping cache fresh but adding write latency.

What is TTL (time to live) in caching and what trade-off does it control?

TTL is the duration a cached item remains valid before expiring. It trades freshness against load: short TTL = fresher data but more origin hits; long TTL = fewer origin hits but staler data.

What is Cloud CDN and what does it cache?

Cloud CDN uses Google's global edge network to cache content (typically static/cacheable HTTP responses) close to users, reducing latency and backend load; it works with the external HTTP(S) Load Balancer.

What is Memorystore in Google Cloud and what is it used for?

Memorystore is a fully managed in-memory data store service for Redis and Memcached, used for low-latency caching and session storage.

What is graceful degradation as a failure-handling strategy?

Designing a system to continue providing reduced/partial functionality (e.g., serving cached or default content) when a dependency fails, instead of failing completely.

What is autohealing in a Managed Instance Group?

A self-healing mechanism where the MIG uses application-level health checks to detect and automatically recreate (replace) instances that are unresponsive or failing.

What is the bulkhead pattern in resilient system design?

Isolating resources (e.g., connection pools, thread pools) per dependency or tenant so that a failure or overload in one part cannot exhaust resources and cascade to take down the whole system.

What is a dead-letter queue (topic) in Pub/Sub and why is it used?

A separate topic where messages that repeatedly fail processing (exceed max delivery attempts) are routed, preventing them from blocking the subscription and allowing later inspection/reprocessing.

What is the difference between MTTR and MTBF in reliability operations?

MTBF (Mean Time Between Failures) measures average uptime between failures (reliability); MTTR (Mean Time To Repair/Recover) measures average time to restore service after a failure (recoverability).

What is a canary deployment and how does it reduce risk?

A release strategy that routes a small percentage of traffic to the new version first, monitoring it for errors before gradually shifting all traffic—limiting blast radius if the new version is faulty.

What is the difference between a blue-green deployment and a rolling update?

Blue-green keeps two full environments and switches traffic from old (blue) to new (green) at once, enabling instant rollback. A rolling update incrementally replaces instances of the old version with the new one, with no second full environment.

What is a 'golden signal' in SRE monitoring, and what are the four golden signals?

Golden signals are the key metrics to monitor a user-facing system: Latency, Traffic, Errors, and Saturation.

What is log-based metrics in Cloud Logging?

Metrics derived from the content of log entries (counter or distribution types), letting you create metrics and alerts from log data—e.g., counting occurrences of a specific error message.