Server Health
Loading…
Build & Capabilities
Loading…
Collections
Loading…
Cache Effectiveness
Rate per refresh interval (delta of cumulative counters). Accumulates while the page polls.
Hits by tier (l1 hot → l3 cold)
no cache metrics
Object-store read pressure (TD-IOTRACE-2 physical KRU)
Ranged-GETs and bytes read per refresh interval.
Ops by operation
no object-store metrics
Write bytes by access tier (hot → archive)
no write-by-tier metrics
Network / embedding outgress
no outgress metrics
KOU = bytes moved out of the deployment (Dimension 2). KEU = embedding units (Dimension 5). Neutral telemetry; pricing lives in the control plane.
Request latency (HTTP)
no latency histogram
p95 over time
Quantiles computed from the
http_request_duration_seconds histogram buckets (aggregated across routes). Status color vs 100 ms / 500 ms thresholds is indicative.Compute time by engine
no task-execution metrics
SQL conformance
TPC-H (22) and TPC-DS (16) suites are the conformance driver and run as ratcheted
qa-gate integration suites — their pass counts are not emitted as live Prometheus series.
Track progress via the qa-gate runs; each newly-passing query is the ratchet of progress toward
full ANSI SQL over pgwire.
Live query-shape counters (e.g.
proximadb_task_execution_time_ms) appear left when present.io_trace sink (TD-TRACE-2)
no io_trace metrics — feature off or warehouse unconfigured
The Iceberg-managed Parquet warehouse sink ships query-scoped I/O traces. Records dropped / upload failures are colored when non-zero. Requires the
io-trace feature and a configured warehouse.Physical ranged reads (per refresh)
Same physical GET / bytes-read counters surfaced in the Storage tab, plotted over time.