Context

The OpenShift Logging console UI has the capability to showing users a histogram of the logs ingested, users can press the "Show histogram" button to display this histogram. In order to display this histogram the UI has to perform a histogram query to Loki for the time window that users are requesting. Histogram queries are expensive so inherently two things will happen for customers consuming large amounts of logs:

• The histogram query will fail with a timeout and users will never see the histogram
• The query will not fail but the memory used by queriers will spike possibly leading to their eviction.

Both are not a great experience for the uninformed user that might press the button to visualize their logs.

Nowadays Loki provides the APIs that allows us to perform queries about the logs ingested without having to load into memory all the logs (these APIs query only TSDB data). Query Log statistics API or Query log Volume.

Goal

The goal of this spike would be to see how could we make the press of "Show histogram" button more performant through the APIs mentioned before.

Developer notes

• Some initial investigation seems to suggest that nowadays none of the two APIs seem to gives directly the answer we want. We might need to make a contribution upstream to create GET /loki/api/v1/index/stats_range