This is a good one, among several bug fixes, we have batched Series and Query RPCs now - this should lead to drastic improvements in network bandwidth and decent improvements in CPU/Memory usage of Querier and all its endpoints.
Changes
Fixed
- #8378: Store: fix the reuse of dirty posting slices
- #8558: Query-Frontend: Fix not logging requests when external-prefix is set in query
- #8254 Receive: Endless loop of retried replication with capnproto and distributors
- #8618: Query-Frontend: ensure slow query and stats logging on request failures
- #8480: Store: fix(readerpool): avoid inserting nil LazyBinaryReader on error to avoid panic
Added
- #8623: Query: Enable batching of Series per SeriesResponse.
- #8582: Sidecar: support --storage.tsdb.delay-compact-file.path Prometheus flag.
- #8595: *: add --shipper.upload-compacted flag for controlling upload concurrency in components that use shippper
Changed
- #8630: Receive: breaking ⚠️ shuffle sharding now uses consistent hashing for stability on scale. Existing tenant-to-node assignments will change on upgrade.
- #8555: Promu: re-add Darwin and FreeBSD as release platforms
- #8661: Upgrade Prometheus to v3.8.0.
tsdb.enable-native-histogramsflag in Receiver and Ruler becomes a no-op and deprecated. Now native histogram ingestion is always enabled.
New Contributors
- @hek2mgl made their first contribution in #8425
- @jrisebor made their first contribution in #8569
- @prymitive made their first contribution in #8582
- @abhiweye made their first contribution in #8592
- @TessaIO made their first contribution in #8556
- @dethi made their first contribution in #8622
- @adezxc made their first contribution in #8623
- @erlan-z made their first contribution in #8480
- @dowgird made their first contribution in #8595
- @pvlltvk made their first contribution in #8618