Precompiled 2024-06-10-dc316ab

Strip unused dependencies based on cargo machete (just machete) (#26867)

Remove bunch of unused dep arrows in the dep graph.

GitOrigin-RevId: e6d9c293231dadcbbba241b40519614f3058103c

Precompiled 2024-06-07-00f5ec1

run SchemaWorker in all namespaces (#26817)

run schema worker in every namespace that has a `_schemas` table

GitOrigin-RevId: e1a9776a68c7c9bed8300f682c62f41c30bc28f5

Precompiled 2024-06-04-17f0813

Add bigbrain support for searchlight clusters (#26562)

GitOrigin-RevId: 692cb9394db34b42015f5f784868086ebe7c651b

Precompiled 2024-06-03-2f0de6f

[Document Retention] Hash on `id` not `timestamp` and `id` (#26486)

The goal of this partitioning is that all documents with the same `id` are deleted in the same chunk. Except, we were hashing on the `id` and `timestamp`, so documents with the same `id` wouldn't always be in the same chunk. This is bad because we can delete more documents than intended.

Now, we will hash only based off of the document `id`. I added a test and confirmed that this failed with the old hashing method.

GitOrigin-RevId: d00d8692d70253762c9ec6762abe1e2e7934db2e

Precompiled 2024-05-31-8fcd070

Enable tests for multisegment text search path (#26441)

This PR fixes search scoring to work for multisegment search query path.

GitOrigin-RevId: dc744089b870ff3abe6da79738834eed041ca7d2

Precompiled 2024-05-30-c553c3c

Generalize logging in compactor (#26398)

GitOrigin-RevId: 734bbdf093bf243ba9efaad842ea612e384dac83

Precompiled 2024-05-29-e147dad

Remove vector index flusher in favor of factory methods for SearchFlu…

…sher (#26359)

GitOrigin-RevId: eb3f6568ea5280ad6579bf8e61235d4eb22c75be

Precompiled 2024-05-24-18c00df

Sync Worker: Always compute transition at latest timestamp (#26258)

I am converting sync worker to work with Usher, and if I keep the code as is, we will make multiple calls to latest_timestamp() per transition. One to determined the target timestamp and one to check it.

Making update_scheduled a Option<Timestamp> is generally a good idea, but we never actually leveraged it. So it is easier to revert for now. I have sent a separate RFC on how to finish the swing, but this provides little benefit until we have sync workers running on the edge.

GitOrigin-RevId: 8b40ebb2d5c83b50dd7ed8cac26f6da8b088cc29

Precompiled 2024-05-23-637c1d4

TermMetadata RPC (#26203)

This PR adds a new trait `TermMetadataFetcher` that will get the term metadata for a segment given term values and the number of documents containing that term that were deleted. This PR doesn't hook it up to the `build_new_segment` code, but the end goal is to get rid of `TermDictionary` lookups in backend. The searcher downloads the relevant segment, gets term ordinals based on the term value, and counts the number of terms that have been completely deleted (num_deleted_docs = doc frequency in the segment).

GitOrigin-RevId: f911381c2b17272575c1bc672343e15d5c5f2723

Precompiled 2024-05-22-48e3aeb

Remove unused dashboard deps (#26006)

GitOrigin-RevId: dfe1cc15c967555c05191c2b1a70ed99a2573c65