Dynamic metric aggregation

[nvartolomei]

In Seastar, metrics can be defined with implicit aggregation by specific labels, which occurs at query time. This feature is useful, for instance to define metrics per shard or even more finely grained per an application-defined entity while reporting them in a more aggregated manner, such as sum or histogram per server.

However, there are times when it is necessary to inspect the fine-grained metrics. I'm sending a PR to allow disabling this aggregation and get the "raw" metrics by adding __aggregate__=false to the query string. https://github.com/scylladb/seastar/pull/2219

---

Separately, what do we think about adding dynamic aggregation capabilities.

E.g.

• aggregate[query_latency]=false to disable aggregation just for query_latency metric
• aggregate[query_latency]=-table to remove table label from aggregation; i.e. report metric per database table
• etc.

Thoughts?

@amnonh @StephanDollberg