Datadog Quiz
Last Updated: September 13, 2026
What remains the team's responsibility with Datadog SaaS?
- A) Nothing after installing the Agent
- B) Only selecting a dashboard color
- C) Collectors, identity, instrumentation, data handling, monitors and cost
- D) Datadog's physical database servers
Show Answer
Answer: C
SaaS manages the backend. APM, profiling, logs and other products have distinct entitlements and billing; an Agent does not include everything.
Which credential/integration statement is correct?
- A) Baseline Agent ingestion needs an API key; application keys and AWS account roles serve additional, specific features
- B) Every Agent needs an application key and broad AWS read role
- C) Adding IRSA automatically configures Datadog SaaS AWS integration
- D) A guessed service-account name is sufficient
Show Answer
Answer: A
The external metrics provider needs additional API permissions/key configuration. SaaS AWS integration uses an authorized cross-account role/external ID. Resolve the actual rendered Agent SA.
Does admission.datadoghq.com/enabled=true alone prove APM SDK injection?
- A) Yes, including every language/version automatically
- B) No; configure SDK annotations or SSI targets, then verify newly admitted pods and actual trace data
- C) Yes, even in the Cluster Agent namespace
- D) Yes, if a trace socket exists
Show Answer
Answer: B
Mutation/connection settings and library injection are distinct. Current local injection excludes kube-system and the Cluster Agent namespace. Library, runtime, mount and security compatibility still matter.
How should an application pod reach the node DogStatsD Agent?
- A) Always use the application's localhost
- B) Put the API key in every UDP packet
- C) Create an unrelated ConfigMap
- D) Use the configured reachable endpoint, such as a mounted Linux UDS directory
Show Answer
Answer: D
The application localhost is not a node Agent. UDS paths, permissions and SDK argument formats must match. Datagrams do not acknowledge SaaS ingestion; counters are not an exactly-once ledger.
Which metric interpretation is correct?
- A) kubernetes.cpu.usage.total is percent
- B) All missing legacy-catalogue metrics were removed
- C) kubernetes.cpu.usage.total is nanocores; Kubelet restart metrics are cumulative gauges
- D) Summing repeated restart samples counts new restarts
Show Answer
Answer: C
system.cpu.idle is percent. Kubelet and State Core have different valid metric names and tags. The example restart monitor explicitly evaluates a total; recent increases need reset-aware validation.
What does the .as_count() error-ratio path calculate?
- A) The ratio of time-aggregated error and total counts
- B) A sum of every time-bucket ratio
- C) A global p95
- D) Automatic 100% success for zero traffic
Show Answer
Answer: A
Use sum aggregation and matching groups. The helper emits zero good/error counts explicitly. Zero traffic, missing data and error-free traffic remain different states.
Which OpenMetrics/log configuration statement is correct?
- A) Any ConfigMap is automatically mounted
- B) Use matching container annotations/current check fields; Logs Grok rules use match_rules/support_rules
- C) prometheus.enabled at chart root configures everything
- D) Grok camelCase keys and snake_case keys are equivalent
Show Answer
Answer: B
The current OpenMetrics check uses openmetrics_endpoint. datadog.confd provides chart-owned mounting; standalone ConfigMaps do not self-install. Request-schema validation is not a live scrape or Grok parse.
What should manual trace-log correlation preserve?
- A) Only dd.trace_id, deleting all other MDC fields
- B) An arbitrary numeric cast of a 128-bit ID
- C) A hardcoded successful trace ID
- D) The caller's prior MDC context, string IDs and actual instrumentation/data prerequisites
Show Answer
Answer: D
The helper restores context even when application code raises. It is synchronous. Automatic injection/parsing, consistent service tags and available traces are separate requirements.
What is wrong with pricing 50 services as 50 APM hosts?
- A) APM is always free
- B) Log ingestion is the whole log bill
- C) Services and billable hosts are different units; product/contract allotments and usage must be counted
- D) Every cluster has one host
Show Answer
Answer: C
The old estimate was not a measured bill. Indexing/retention, span allotments, custom metrics and other products matter. nonLocalTraffic is reachability, not a cost quota.
- Which Watchdog/SLO/diagnostic practice is correct?
- A) A Watchdog insight proves a page was delivered
- B) Match the SLO model and good/total policy, test routing, and inspect local diagnostic bundles before sharing
- C) A local flare automatically authorizes upload
- D) Dump every DD_ environment value when traces are missing
Show Answer
Answer: B
Datadog supports metric-, monitor- and time-slice SLOs. Notification and no-data behavior need validation. env dumps can expose keys; --local keeps the initial flare collection local.