Cyril Tovena
Cyril Tovena
There could be a lot of reason. If there is only few queriers that are doing work then possibly I'll be looking at increase or decreasing the worker config for...
What's the query stats ? The max_concurrent should reflect more cpu allocated per querier. In query range you should activate sharding.
I'm following closely this PR. As previous member state, this is not the direction we want, but we still want to make Loki faster. To help us ensure we'll cover...
how do you deploy Loki ? for 10TB/day you need to have the distributed deployment.
Can you share statistics of a query?
Let's see if we get more traction, I like the mutable idea 👍 . I expect more people to want this at high scale to avoid allocations but if this...
I can confirm this can be useful when running a custom path prefix. Pretty much likes Prometheus scraping path.
Latest progress: ``` ❯ go test -timeout 0s -run ^TestCompact$ github.com/grafana/phlare/pkg/phlaredb -v === RUN TestCompact compact_test.go:70: src block(# 0 ) minTime 2023-07-12T03:18:05.463+02:00 maxTime 2023-07-12T06:18:19.522+02:00 numSeries 5785 numProfiles 2650992 numSamples 168740499...
Local compaction (download blocks first) takes 1m20. https://flamegraph.com/share/1f92c67b-208c-11ee-85f3-ba195a7e5c95
We just need to rewrite stacktraceids and ingest symbols and then add the files meta and that should be good to go.