Multiple metrics for the same connection.

[tronda]

We have an ASP.NET Core service which sends metrics to an OpenTelemetry Collector which again exports those metrics to Prometheus. With this setup we see that the service is reporting the same metrics values three times where the only difference is the content of the instance label:

odp_number_of_pooled_connections{Instance="MyService[1,1][Data Source=med-devdb.dev.dips.no:1521/mydb;Min Pool Size=1;Max Poo...][mydb]", container="metrics-node", ...
odp_number_of_pooled_connections{Instance="MyService[1,1][Data Source=med-devdb.dev.dips.no:1521/mydb;Min Pool Size=1;Max Poo...]", container="metrics-node", ...
odp_number_of_pooled_connections{Instance="MyService[1,1]", container="metrics-node", ...

The challenge is when we create a dashboard and the content of the Instance is the only thing that separate those metrics, it is a bit challenging to create a good dashboard since we really is only interested in one of this metrics.
Is it possible to change the behavior so that only one metric will be reported? If this is not possible, should there be a label/tag in the metric which can be used to filter out the one metric that we care about without doing text content check?

[alexkeh]

Hi @tronda,
There isn't a way to filter for one instance's metrics at the ODP.NET level.

The metrics are reported at the app domain, pool, and DB instance levels, which is the reason why you see these three entries. Having multiple levels can be helpful if you are using a RAC DB or other multi-instance setup.

This is the same model ODP.NET uses with Windows performance counters. Windows Performance Monitor users can select the "metric" they want to monitor. They can filter out one or more of these ODP.NET counters.

Can you filter for one of the ODP.NET metrics strings? For example, only show entries with "[mydb]" in the filter.

[tronda]

Hey Alex, and thank you for the feedback.

I would suggest that Oracle add a label (tag in System.Diagnostics.Metrics API) which indicates if the metric is either app domain, pool or db-instance because this will give a much more efficient filtering.

We are having a Grafana dashboards where we look at hundreds of microservices which all report these metrics and doing text filter is not a scalable way of doing this.

[alexkeh]

Thanks for the suggestion @tronda! Let me share it with my dev team.

[alexkeh]

I have filed ER 38706806 to track this request.