Temporal metric for task queue size/backlog, or schedule to start latency for task queue

Community Support
1

I’m trying to understand which metric I can look at to understand the “backlog size” of a given task queue. I found Get the count of currently active workflows and activities - #3 by Michael_Mole, though this post was discussing how to retrieve this count programmatically, which I’m not interested in - simply the metric.

If that’s not supported, looking at the scheduled_to_start latency metric is also acceptable to understand how “behind” we are for a given queue. I can see this is defined as “task_schedule_to_start_latency” from the Metrics defs.go. However, I’m having trouble finding this for a given queue. Here’s what I can see today, by doing a search in DataDog:

Screen Shot 2021-12-01 at 7.15.57 PM
Screen Shot 2021-12-01 at 7.15.57 PM1444×204 52.4 KB

None of these metrics appear to contain values related to our specific workflows or task queues. Is there a different metric that allows us to understand schedule-to-start for a given task queue?

2

Schedule to start is a client-side metric. You have to enable it on the client. See this Go sample that sets up Prometheus metrics scope in the worker. Let us know if that helps.

3

Thank you, this was successful. We are using the Java SDK so we followed the discussion in this thread: How to enable Metrics Support using JAVA SDK

4

Glad to hear! We are also adding similar sdk metrics sample for java sdk, see pr here.