Google Cloud Monitoring exponential histogram example
This example shows how to use go.opentelemetry.io/otel to instrument a simple Go application with metrics and export the metrics to Google Cloud Monitoring
Example details
This example simulates server latency by sampling a Log-Normal Distribution with the parameters $\mu = 3.5 , 5.5$ and $\sigma = .5$. We generate the following three example of server latency :
- Latency with Log-Normal Distribution.
- Latency with Shifted Mean Distribution.
- Latency with Multimodal Distribution (mixture of 1. and 2.).
We explore the resulting distributions with three types of histograms :
- Opentelemetry Default Linear Buckets Histogram. Buckets are
0, 5, 10, 25, 50, 75, 100, 250, 500, 1000
- Opentelemetry Linear Buckets Histogram. Buckets are
0, 10, 20, 30, ..., 340, 350.
- Opentelemetry Exponential Buckets Histogram with default parameters.
Build and run the application
Go to the example/metric/README.md instructions that describe how to build and run all examples.
Sample execution
$ ./metrics
2024/03/21 15:38:58 Sent Latency Data (Original Distribution): #points 1000 , mean 36.64255895183214, sdv 19.670797833645373
2024/03/21 15:38:58 Sent Latency Data (Shifted Distribution): #points 1000 , mean 277.70002931783233, sdv 143.59582355437485
2024/03/21 15:38:58 Sent Latency Data (Multimodal Distribution): #points 1000 , mean 151.49111863163805, sdv 159.2187295223318
...
Create dashboard
When filling in the Find resource type and metric box, use the metric names with the prefix "workload.googleapis.com/latency_" to observe histogram metrics (for example "workload.googleapis.com/latency_a").
If you already know how to use Cloud Monitoring and would just like to confirm the data is properly received, you can run the dashboard creation script bundled in this directory. This command requires at least the roles/monitoring.dashboardEditor permissions to create a new dashboard.
$ ./create_dashboard.sh
This script creates a dashboard titled "OpenTelemetry - Exponential Histogram example".
You should be able to view histogram charts like below once you create the dashboard.