Is Chunks Cache Required if Running Blocks Storage Engine?

[michael-burt]

Is the Chunks Cache, documented here, a component that should be deployed when using the Blocks storage engine?

This page includes documentation for Index Read, Index Write, Query Results, & Chunks Caches. Are all of these caches irrelevant for Blocks storage engine? I see both the Index Read & Chunks Cache on the Blocks Storage architecture diagram here.

[dgonzalezruiz]

Yes, although the wording can be confusing. From https://cortexmetrics.io/docs/blocks-storage/ -->

). Each block is composed by chunk files - containing the timestamp-value pairs for multiple series - and an index, which indexes metric names and labels to time series in the chunk files.

So, for caching your blocks storage query path / avoiding to refetch already queried samples, you would have to at least run both chunk and index cache.
You also can configure a separate cache for metadata & query results for even more caching, and there are some more tips for scaling it on https://cortexmetrics.io/docs/blocks-storage/production-tips/#caching

[michael-burt]

Thanks for the explanation @dgonzalezruiz