Thanos query frontend caching for metadata API's

[krishnaindani]

Currently Thanos query frontend cache and split /api/v1/query_range calls as it says in query frontend docs. Does query frontend cache metadata as well? I see the parameters prefixed with --labels mentions retrieving labels through Labels and Series API, but I am confused with what it does? For example does, --labels.response-cache-config is the cache config for series and labels API?

The use case I am looking for if we have a Thanos query frontend which queries thanos query which in turns fans out to multi store and sidecars, now when we use explore tab in Grafana the query from the Metric browser is really expensive and querying high cardinality metrics the thanos and prometheus crashes. The sum of metrics series(through scrape_samples_post_metric_relabeling) across our pipeline is 26M.

I have also posted in slack and this issue is pressing us.

[krishnaindani]

would someone be able to guide on this?