OpenMetrics FAQ
Which Cloud metrics and labels are available?
OpenMetrics includes metrics for Activity outcomes and latency, Task Queue backlog, Namespace usage and limits, service requests, and Workflow outcomes. Metrics can include labels such as region, temporal_task_queue, and temporal_workflow_type. Some high-cardinality labels, including temporal_activity_type, require you to opt in when scraping.
For the current list of metric names, descriptions, and labels, see the OpenMetrics metrics reference.
Can I query Cloud metrics directly?
The OpenMetrics endpoint returns the most recently completed aggregation window for scraping. It does not provide a time-series query API. Configure an observability tool to collect and retain the metrics, then query them in that tool.
How often can I scrape, and how much data can I retrieve?
See OpenMetrics API limits for the request and datapoint limits. If a scrape exceeds the datapoint limit, use Namespace or metric filtering to reduce the response size.
Why are some metrics missing from a scrape?
The OpenMetrics endpoint returns only metrics reported during its one-minute aggregation window. For example, temporal_cloud_v1_workflow_failed_count is absent if no Workflows failed in that window. A Task Queue with no activity can also have no metrics in a scrape. The set of returned metrics can therefore vary between scrapes.
If a dashboard or alert should treat an absent series as zero, handle that in your observability tool. For example, in Prometheus:
sum(temporal_cloud_v1_workflow_failed_count) or vector(0)