Grafana Tempo Quiz
Last Updated: September 13, 2026
Baseline: Tempo 3.0.3 and chart 3.6.0.
What best describes Tempo storage and search?
- A) Every attribute must be indexed in Elasticsearch
- B) Object-store Parquet blocks support TraceID/TraceQL; storage and querying still have costs
- C) Knowing an ID recovers every dropped span
- D) Tempo retains traces forever
Show Answer
Answer: B) Object-store Parquet blocks support TraceID/TraceQL; storage and querying still have costs
Dedicated columns, metadata and caches do not mean zero indexing or query cost. Only successfully ingested, retained data is available.
Which component receives and validates trace data before the Tempo 3 distributed write path commits to Kafka?
- A) Block-builder
- B) Querier
- C) Distributor
- D) Backend worker
Show Answer
Answer: C) Distributor
The distributor writes to Kafka. Live-stores, block-builders and optional metrics-generators consume separately; this is not the Tempo 2 ingester path.
Which TraceQL query selects spans with error status?
- A)
{ duration > 1s } - B)
{ status = error } - C)
{ status = ok } - D)
{ span.http.response.status_code = 200 }
- A)
Show Answer
Answer: B) { status = error }
Span status error is distinct from a latency threshold or an arbitrary HTTP response condition.
Which identity configuration does this EKS/S3 example use?
- A) Static access keys in Helm values
- B) The node role shared by every workload
- C) IRSA bound to monitoring:tempo with exact OIDC sub/aud and scoped S3 permissions
- D) An unrelated ServiceAccount with no Pod association
Show Answer
Answer: C) IRSA bound to monitoring:tempo with exact OIDC sub/aud and scoped S3 permissions
The role annotation and every Tempo Pod ServiceAccount must agree. Other workload-identity approaches need their own pinned-image compatibility check.
Which is not generated from traces by the illustrated metrics-generator processors?
- A) Service graph metrics
- B) Span metrics
- C) Arbitrary application log metrics
- D) Rate/error/duration metrics derived from spans
Show Answer
Answer: C) Arbitrary application log metrics
Span-metrics and service-graphs need explicit processor activation and remote write. They do not turn arbitrary logs into metrics.
Which statement about Tempo 3 durability is correct?
- A) Three Tempo replicas always guarantee zero loss
- B) Microservices use Kafka; its replication, ISR, retention and recovery must be designed separately
- C) Monolithic mode always requires Kafka
- D) Every StatefulSet automatically has a persistent PVC
Show Answer
Answer: B) Microservices use Kafka; its replication, ISR, retention and recovery must be designed separately
The chart uses emptyDir for live-store/block-builder data. Kafka durability is not established by a Tempo replica count; monolithic mode does not require Kafka.
Which Grafana correlation direction is correct?
- A) Same namespace alone creates correlation
- B) Tempo tracesToLogsV2 provides Trace→Logs; Loki derivedFields provides Logs→Trace
- C) Both systems must share an S3 bucket
- D) derivedFields causes applications to generate TraceIDs
Show Answer
Answer: B) Tempo tracesToLogsV2 provides Trace→Logs; Loki derivedFields provides Logs→Trace
Identifiers, data-source UIDs, labels and the queried time range must match real data. A link cannot recover missing telemetry.
Which components handle Tempo 3 background compaction and retention work?
- A) Grafana browser tabs
- B) OTLP clients
- C) Backend scheduler and backend workers
- D) The old compactor configuration copied unchanged
Show Answer
Answer: C) Backend scheduler and backend workers
These replace the old compactor architecture. Retention is asynchronous, and an independent blanket S3 expiration rule can conflict with backend operations.
What does
{ resource.service.name = "A" } >> { resource.service.name = "B" }select?- A) Any two spans in different traces
- B) Matching B descendants of matching A spans
- C) Only A parents, never B
- D) Only direct B children
Show Answer
Answer: B) Matching B descendants of matching A spans
The result is on the right-hand side. Use > for direct children. Neither sibling matching nor a same-trace membership test means the same thing.
- What is the safest first response to slow queries and apparently empty recent searches?
- A) Copy ingester.max_block_duration: 30m from Tempo 2
- B) Disable all lag and recent-query protections
- C) Check time range, actual received data, lag, scan volume and limits before tuning
- D) Turn missing telemetry and zero traffic into a guaranteed healthy value
Show Answer
Answer: C) Check time range, actual received data, lag, scan volume and limits before tuning
Tempo 3 has different components/defaults. Empty results, zero traffic and failures are distinct; configuration rendering alone does not prove a production system works.