Metrics and alerts
The metrics e6data exposes, the dashboards worth building, and the alerts to set on your observability stack.
e6data exposes operational telemetry that you route to your own observability stack via OpenTelemetry - see Observability export for setup. This page covers what to dashboard and which alerts to set.
Key metrics
Cluster health
Executor count, CPU and memory, queue depth
Query performance
P50/P95/P99 query latency, slow-query count
Workload
Queries per minute, concurrent queries, cache hit rate
Errors
Failed-query rate, error-type breakdown, timeout rate
Dashboards to build
Cluster health - executor count, CPU/memory, and queue depth per cluster.
Query performance - latency percentiles and slow-query count over time.
Workload patterns - throughput, concurrency, and cache hit rate.
Errors - failed-query rate and the breakdown by error type.
If your observability backend has prebuilt e6data dashboards, start from those.
Alerts to set
Cluster fails to resume
More than 5 minutes in Pending after a resume is triggered
Failed-query rate spike
Above your established baseline for the cluster
Query-timeout spike
A sudden increase in timed-out queries
P99 latency breach
Above your latency threshold, per cluster
Cache hit rate drop
Below the expected baseline
Workspace disabled unexpectedly
Any unplanned workspace disable
Set these in your observability backend on the exported metrics; e6data doesn't send alerts itself.
See also
Last updated
Was this helpful?