Bucket bytes download metric by filetype

[jlsherrill]

Is your feature request related to a problem? Please describe.
Currently the content type emits download metrics that feature bytes downloaded.

For the rpm plugin, we would like to know what is costing more bandwidth: metadata or rpms.

Describe the solution you'd like

One solution would be to bucket the metrics by filetype (using the extension of the file), so we can query just the rpms:

sum by(domain_name) (rate(pulp_artifacts_size_counter_Bytes_total{filetype="rpm"}[2m]) * 60)

Describe alternatives you've considered

Additional context

[dkliban]

@jlsherrill What's the difference between this and #5815

[lubosmj]

This one is a feature request. It would be nice if we could determine which file types are being streamed through content-app and record this inside the metric. The other one is trying to fix/verify the current workflow of the existing metric.

Actually, I had in mind another issue. I think these are duplicates. We should close one of them.

[jlsherrill]

This was for the existing metric related to download bytes (how much data was downloaded), #5815 is for a new metric to measure the number of files downloaded.

[dkliban]

We will add an additional attribute to the existing metric. The attribute will describe the type of file served: rpm or metadata.

[lubosmj]

Nope. The attribute will describe the file extension.

It is plausible that we will extract this information from the header:

pulpcore/pulpcore/content/handler.py

Line 392 in dfa6dfb

headers["Content-Type"] = content_type

.

[lubosmj]

We decided to record the file extension only when the HTTP response code is equal to 3XX or 2XX.

[lubosmj]

I have just realized that we do not emit the metric in case the caching is disabled... 🎱