Some improvements ....

[carlozleite]

I'm using metrics-exporter to scan all azure resource providers and I noticed that in some cases the values ​​returned null, this generated an exception, so I put two "if's" to check if these data were null, avoiding the exception .

Another change is related to labels that are not being inserted correctly, generating the exception "was collected before with the same name and label values", also adds a new label called "resource_type" referencing the name of the resource provider.

I also inserted some string formatting, because some metrics incline the character "-" this caused failure in prometheus "no token found" ...

And lastly I added a part of the resource provider name to the metric name, removing only the name "Microsoft" to make it easier to identify within prometheus.

Example:

Compute_percentage_cpu {instance = "Azure-Compute: 9090", job = "Azure-Compute", resource_group = "Development", resource_name = "vmdtscience", resource_type = "Microsoft.Compute"}

[conr]

Thanks for improving the azure-metrics-exporter!

Can you explain what your improvements are? :)

[carlozleite]

Hi Conor, I updated the description of the PR with the changes.

[carlozleite]

On my fork I added support for Docker ... if you are interested only change the path of the repository in the docker-build.sh and Dockerfile files.

[conr]

I'll try get to testing this sometime this week.

Would you be able to separate this into two separate PRs?
One for the docker stuff and one of the code changes.

Thanks!