SD-11577: add support for kv metrics

[rtalvarez]

Pull Request Template

Description

Adds Cloudflare KV storage metrics to the Prometheus exporter. Two new metrics are exposed:

• cloudflare_kv_requests_count — number of KV operations by namespace and action type (read/write/delete/list)
• cloudflare_kv_latency — KV operation latency quantiles (P50/P75/P99/P999) in milliseconds

Namespace IDs are resolved to human-readable names via the Cloudflare REST API (/storage/kv/namespaces), falling back to the raw ID if the name cannot be resolved.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Code refactoring
• Other (please describe):

Testing

Ran the app locally and verified that cloudflare_kv_latency and cloudflare_kv_requests_count are visible by visiting /metrics

Verified the graphql query gets executed by the app

DEBU[2026-03-17 15:47:25] func:main.NewGraphQLClient.func1 file:graphql.go >> query:
        query ($accountID: String!, $mintime: Time!, $maxtime: Time!, $limit: Int!) {
                viewer {
                        accounts(filter: {accountTag: $accountID}) {
                                kvOperationsAdaptiveGroups(limit: $limit, filter: {datetime_geq: $mintime, datetime_lt: $maxtime}) {
                                        dimensions {
                                                namespaceId
                                                actionType
                                        }
                                        sum {
                                                requests
                                        }
                                        quantiles {
                                                latencyMsP50
                                                latencyMsP75
                                                latencyMsP99
                                                latencyMsP999
                                        }
                                }
                        }
                }
        }

• I have run make pr-tests and all tests pass
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

Code Quality

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation

Before Submitting

Please ensure you have completed the following before submitting your PR:

# Run comprehensive tests
make pr-tests

If the above command fails, please fix the issues before submitting your PR.

Additional Notes

Add any other context about the pull request here.

[chanceaclark]

🍹 From experience, this list is going to get longer and longer the more native hosting stores we have. I think it might be fine for now, but for example, we do the auto paging for custom hostnames and it takes ~5 minutes to get through 90k custom hostnames.

[rtalvarez]

hmm you're right. do you think it's needed to fetch KV namespaces on every scrape? i think it might be a bit overkill. maybe we can cache them on boot and then refresh the cache on a separate interval (different from the scrape interval)

[chanceaclark]

1. Only add namespace_name if the id matches a hardcoded list - need mapping table.
2. Only add namespace_id if the id matches a hardcoded list - simple slice.
3. Add both namespace_name and namespace_id if id matches a hardcoded list – need mapping table.

All other namespaces that are not on the hardcoded list don't have the label because it's high-cardinality.

[rtalvarez]

added changes to implement 2) in e3fb3b0

a quick caveat: prometheus constraint is that the same set of labels have to be set on every series, we can't omit the namespace_id entirely. so i've defaulted it to other so they all get bucketed there.