▸case-07 Filtered vector searches in Qdrant are taking 500ms despite low vector count. Developers suspect HNSW graph depth and plan to lower 'ef_construct'. Which request duration metrics and collection index checks reveal whether the slowdown is caused by unindexed payload field scans during filtering? | pass→pass | 16,200 | 11,463 | -29% | 1 | 1 | 0% | 2,787 | 1,916 | -31% | 0 | 0 | — |
▸case-08 Under peak write load, client applications sending batch updates to Qdrant experience timeouts. Engineers consider increasing HTTP request timeouts. What write-ahead log (WAL) and background queue metrics should be checked to identify write backpressure? | pass→pass | 15,185 | 14,408 | -5% | 1 | 1 | 0% | 2,401 | 2,507 | +4% | 0 | 0 | — |
▸case-01 We are configuring Prometheus to pull operational telemetry from a Qdrant cluster node running on standard ports. Engineers suggest scraping the root HTTP path '/' or port 6334. What specific HTTP endpoint path and default REST port should be configured in our Prometheus scrape target job? | fail→pass | 11,204 | 7,065 | -37% | 1 | 1 | 0% | 620 | 814 | +31% | 0 | 0 | — |
▸case-02 We are setting up Kubernetes liveness probes for a Qdrant pod deployment. A developer proposed probing the gRPC port or using '/ping'. Which HTTP REST health probe path and expected HTTP success code should be used in the liveness probe spec? | pass→pass | 10,516 | 7,937 | -25% | 1 | 1 | 0% | 1,781 | 1,617 | -9% | 0 | 0 | — |
▸case-09 We need to author a Prometheus alerting rule in PromQL for Qdrant to trigger when HTTP 5xx server errors exceed 5% of total REST requests over 5 minutes. The team proposed matching standard ingress metrics. What Qdrant-native REST histogram/counter metric and PromQL rate query should be used? | fail→pass | 12,561 | 9,318 | -26% | 1 | 1 | 0% | 2,483 | 1,922 | -23% | 0 | 0 | — |
▸case-03 We want Kubernetes readiness probes to hold traffic until Qdrant finishes loading collection segments and WAL logs into RAM. Someone suggested using the liveness probe endpoint '/healthz' for readiness as well. Which specific readiness health check endpoint path ensures traffic is deferred until storage initialization completes? | pass→pass | 7,269 | 5,418 | -25% | 1 | 1 | 0% | 1,276 | 1,105 | -13% | 0 | 0 | — |
▸case-04 For deep internal diagnostic reporting, an SRE wants to parse full structured JSON containing detailed segment counts, vector indices state, and system hardware status from Qdrant. They tried curling '/metrics' but got Prometheus exposition text instead. Which administrative HTTP endpoint outputs this detailed internal diagnostic JSON object? | pass→pass | 8,411 | 5,709 | -32% | 1 | 1 | 0% | 1,329 | 1,072 | -19% | 0 | 0 | — |
▸case-05 During heavy write ingestion in Qdrant, segment merging appears stuck, causing thousands of unmerged delta segments. An operator plans to restart the process immediately. What specific metric or telemetry state should be inspected first to determine whether the segment optimizer is actively processing tasks or starved of threads? | pass→pass | 21,958 | 10,407 | -53% | 1 | 1 | 0% | 3,817 | 1,718 | -55% | 0 | 0 | — |
▸case-06 A Qdrant pod is experiencing persistent RAM growth and approaching container limits. The team assumes adding more container RAM is the only solution. What process memory metric and storage vector configuration option (such as memory-mapping) should be evaluated to mitigate unexpected RAM exhaustion? | pass→pass | 14,504 | 12,774 | -12% | 1 | 1 | 0% | 2,467 | 1,669 | -32% | 0 | 0 | — |
▸case-10 We want a PromQL alert rule that fires when Qdrant 95th percentile search query latency exceeds 100 milliseconds over a 5-minute window. An engineer attempted to use 'avg()' on the duration metric. Which PromQL function and Qdrant request duration bucket metric must be used for p95 calculation? | pass→pass | 8,977 | 12,277 | +37% | 1 | 1 | 0% | 1,756 | 1,974 | +12% | 0 | 0 | — |
▸case-11 When debugging an issue on a Qdrant node, stdout logs are currently formatted as plain unstructured text at info level. SRE wants structured JSON logging with debug verbosity for Fluentbit ingestion. Which environment variables control Qdrant log output verbosity and JSON formatting? | pass→pass | 15,571 | 11,707 | -25% | 1 | 1 | 0% | 3,079 | 1,837 | -40% | 0 | 0 | — |
▸case-12 In a multi-node Qdrant cluster, consensus errors and collection sync delays are occurring across nodes. A developer suggests checking raw ping network response times. What specific Qdrant telemetry or raft consensus status indicators diagnose node peer synchronization and consensus health? | pass→pass | 18,379 | 13,007 | -29% | 1 | 1 | 0% | 2,622 | 2,453 | -6% | 0 | 0 | — |
▸case-13 After millions of point updates and deletes in a Qdrant collection, search latency has degraded even though total point count remains constant. An SRE assumes HNSW graph corruption and wants to rebuild the entire collection. What collection segment metric indicates high segment fragmentation from frequent updates? | fail→pass | 11,115 | 13,511 | +22% | 1 | 1 | 0% | 1,773 | 2,564 | +45% | 0 | 0 | — |
▸case-14 The disk volume housing Qdrant is filling up rapidly due to expanding write-ahead log (WAL) directories. An admin proposes manually deleting WAL files. What operational check determines whether WAL flushing is delayed because vector indexing buffer commits are falling behind? | fail→fail | 14,917 | 13,485 | -10% | 1 | 1 | 0% | 2,414 | 2,220 | -8% | 0 | 0 | — |
▸case-15 In a Kubernetes deployment, Qdrant search latency spikes sporadically while RAM usage stays low. The pod CPU limit is set to 2 cores. Engineers suspect memory swapping. How does CPU throttling under Kubernetes quotas directly impact Qdrant background segment optimizer worker threads? | fail→pass | 16,093 | 15,862 | -1% | 1 | 1 | 0% | 2,636 | 2,727 | +3% | 0 | 0 | — |
▸case-16 We want to write a monitoring script that alerts if any collection in Qdrant degrades from 'green' status to 'yellow' or 'red'. A script developer plans to count points in each collection. Which API endpoint or metric directly provides the collection health status string? | pass→pass | 11,673 | 8,144 | -30% | 1 | 1 | 0% | 1,720 | 1,482 | -14% | 0 | 0 | — |
▸case-17 We support both gRPC and REST client connections to Qdrant and want to build separate Grafana latency panels for each protocol. An engineer assumed a single 'qdrant_request_duration_seconds' metric covers both. What are the specific Prometheus metric names for gRPC versus REST request durations? | fail→fail | 13,109 | 12,454 | -5% | 1 | 1 | 0% | 2,635 | 1,917 | -27% | 0 | 0 | — |
▸case-18 A Qdrant node container keeps getting OOMKilled by the Linux kernel when loading large vectors into RAM. The team considers disabling vector indexing entirely. What container memory alerting threshold and vector storage offloading configuration (such as mmap) should be implemented to prevent OOM events? | pass→pass | 15,692 | 14,032 | -11% | 1 | 1 | 0% | 2,730 | 2,675 | -2% | 0 | 0 | — |
▸case-19 We are building an SRE Grafana dashboard for Qdrant vector database nodes. A developer proposed showing only CPU and disk space percentage panels. What key Qdrant application-level metric panels (such as request latency and memory utilization) must be included for comprehensive observability? | pass→pass | 16,433 | 16,698 | +2% | 1 | 1 | 0% | 2,666 | 2,934 | +10% | 0 | 0 | — |
▸case-20 We need to create a schema payload index on an existing Qdrant collection named 'products' for the field 'category' with 'keyword' field type. An engineer suggests using the snapshot creation endpoint. Provide the correct REST API endpoint and HTTP method to create this payload schema index. | pass→pass | 4,408 | 3,986 | -10% | 1 | 1 | 0% | 677 | 858 | +27% | 0 | 0 | — |
▸case-21 We want to enable Scalar Quantization (SQ) on a Qdrant collection to compress vectors into int8 format while keeping quantized vectors in RAM. Provide the JSON configuration snippet for 'quantization_config' in the collection settings. | pass→pass | 4,264 | 5,173 | +21% | 1 | 1 | 0% | 843 | 847 | +0% | 0 | 0 | — |
▸case-22 We need to trigger an immediate snapshot backup for a Qdrant collection named 'documents' via REST API. An operator wants to use 'GET /collections/documents'. What is the exact HTTP method and endpoint path to request snapshot creation? | pass→pass | 4,002 | 3,193 | -20% | 1 | 1 | 0% | 696 | 701 | +1% | 0 | 0 | — |