promql-to-scrape

README

promql-to-scrape

This basic application is meant to provide an example for how one could use the Temporal Cloud Observability endpoint to expose a typical Prometheus /metrics endpoint.

This example is provided as-is, without support. It is intended as reference material only.

How to Use

Grab your client cert and key and place them at client.crt, tls.key, and your Temporal Cloud account number that has the observability endpoint enabled.

go mod tidy
go build -o promql-to-scrape cmd/promql-to-scrape/main.go
./promql-to-scrape -client-cert client.crt -client-
key tls.key -prom-endpoint https://<account>.tmprl.cloud/prometheus --config-file examples/config.yaml --debug
~~~
time=2023-11-16T17:43:20.260-06:00 level=DEBUG msg="successful metric retrieval" time=3.529039083s

This means you can now hit http://localhost:9001/metrics on your machine and see your metrics.

Important Usability Information

Important: When you go to scrape this, you should do so with a 60s scrape interval, unless you are meaningfully modifying this code. The example queries all assume a 1 minute rate and you'll want these to be equal.

Very Important: The data you will see here is approximately 1 minute delayed (should you conform to the guidance above). Due to the aggregation that happens before metrics are presented to you, it's necessary for us to send the queries from this application to look 60 seconds in the past. Otherwise data aggregation would not be complete, and there would be no results for each query.

Deployment

Some example Kubernetes manifests are provided in the /examples directory. Filling in your certificates and account should get you going pretty quickly.

Generating Config

There is a second binary you can build that can help you build a default configuration of queries to scrape and export.

go build -o genconfig cmd/genconfig/main.go
./genconfig -client-cert client.crt -client-key tls.key -prom-endpoint https://<account>.tmprl.cloud/prometheus 
...

This will generate an example config at config.yaml that you may use. It looks for all the existing metrics and generates a reasonable query for you to export.

• For counters, a rate(counter[1m])
• For gauges, it simply queries for gauge
• For histograms, it does a p99 aggregated by temporal_namespace and operation. histogram_quantile(0.99, sum(rate(metric[1m])) by (le, operation, temporal_namespace)

Modify at your own risk. You may find you'd like to add a global latency across all namespaces for instance. You can add those queries to your config file.