Can you please provide unit of measurement for zeebe_stream_processor_latency_bucket
Is it number of requests in the queue or the time it takes a request to be processed by broker when in queue?
Hey @ankit_joinwal
You can find a description here https://github.com/camunda-cloud/zeebe/blob/develop/engine/src/main/java/io/camunda/zeebe/engine/metrics/StreamProcessorMetrics.java#L47 (which is also exported by the metric endpoint).
And also in the grafana dashboard here zeebe/monitor/grafana/zeebe.json at main · camunda/zeebe · GitHub
Is it number of requests in the queue or the time it takes a request to be processed by broker when in queue?
So to answer your question it is the second.
User commands and internal commands are written to a ring buffer, and get a timestamp assigned. They are then written to the log, replicated and at some point committed. After the command is committed the processor can read and process it. The time it takes from writting to the ring buffer until starting processing is stored in this metric.
Hope that helps.
Greets
Chris
1 Like
Thanks
I see below metrics during our load test
You can see spikes reaching to be 30K value for the latency but then it drops . How do I interpret this graph?
Hey @ankit_joinwal
be aware that this metric is a Prometheus Histogram type, so I’m not sure whether you can show useful data in just a graph view.
This is how we visiualize it in our dashboard, see above for the grafana dashboard (json file).
Please open a new topic/thread if you have further or other questions regarding this.
Greets
Chris