104 Commits

Author SHA1 Message Date
Kerollmops
245a55722a
Remove commented code 2025-02-20 16:48:18 +01:00
Kerollmops
05cc8c650c
Expose the write channel congestion in the batches 2025-02-19 15:47:54 +01:00
meili-bors[bot]
0f1aeb8eaa
Merge #5351
5351: Bring back v1.13.0 changes into main r=irevoire a=Kerollmops

This PR brings back the changes made in v1.13 into the main branch.

Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: Clémentine <clementine@meilisearch.com>
Co-authored-by: meili-bors[bot] <89034592+meili-bors[bot]@users.noreply.github.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
2025-02-18 08:05:02 +00:00
meili-bors[bot]
0c3e7fe963
Merge #5316
5316: Fix the dumpless upgrade corruption r=dureuill a=irevoire

# Pull Request

## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/5280

## What does this PR do?
- Add a test that ensure we write the version in the index-scheduler even if we have a bug while writing the VERSION file
- Do what was described in the issue


Co-authored-by: Tamo <tamo@meilisearch.com>
2025-02-10 09:53:57 +00:00
Tamo
45f843ccb9 fmt 2025-02-10 10:46:42 +01:00
Kerollmops
2b0e17ede0
Make sure arroy is using the rayon thread-pool 2025-02-06 15:28:10 +01:00
meili-bors[bot]
796acd1aee
Merge #5288
5288: Improve AI logging r=dureuill a=Kerollmops

This PR fixes #5285 and brings the changes from #5233 to simplify debugging indexation and search performance issues related to AI. The following texts can be found in the logs to debug and understand performance issues:

 - `embed_one: search` represents the time we spent waiting for the embedding generation, i.e., OpenAI, local HuggingFace, Ollama.
 - `filtered_universe: search::universe` the time spent filtering the documents.
 - ~`next_bucket: search::vector_sort` is the time spent finding the nearest neighbors (ANNs) in the vector store (arroy), locally~ was being triggered too many times.
 - `indexing::vectors` is the time arroy spends indexing the new vectors for a batch.
 - `documents::extract vectors` and `documents::merge vectors` to see the time spent generating and writing the embeddings.

Co-authored-by: Kerollmops <clement@meilisearch.com>
2025-02-04 10:20:45 +00:00
Kerollmops
acc400face
Support merging update and replacement operations 2025-02-03 11:47:17 +01:00
Kerollmops
8e6893ddbe
Make sure we correctly mix different document operations 2025-02-03 10:34:06 +01:00
Kerollmops
424c5bde40
Move the embedding computation and extraction log to debug 2025-01-29 16:40:36 +01:00
Kerollmops
cb1b7513af
Log the memory metrics only once 2025-01-29 15:21:52 +01:00
Kerollmops
db032079d8
Show indexation allocated memory 2025-01-29 14:21:02 +01:00
Clément Renault
a00796c46a
Improve the naming in the log message 2025-01-29 14:21:02 +01:00
Kerollmops
6112bd8caa
Display the channel congestion 2025-01-29 14:21:02 +01:00
Kerollmops
4a5923a55e
log the time arroy took to insert embeddings 2025-01-27 14:22:17 +01:00
Louis Dureuil
f5a4a1c8b2
Give more RAM to bbqueue.
- bbqueue buffers used to have (5% * 2%) / num_threads
- they now have 5% / num_threads
2025-01-24 12:18:32 +01:00
Louis Dureuil
a6470a0c37
Improve error log 2025-01-22 15:50:41 +01:00
Louis Dureuil
8a54f14b8e
Demote panic to error log 2025-01-22 15:49:24 +01:00
Kerollmops
63c8cbae5b
Improve the panic message when deleting an unknown entry 2025-01-14 10:31:44 +01:00
Louis Dureuil
c14967eeac
Use new incremental facet indexing and enable sanity checks in debug 2025-01-09 10:36:35 +01:00
Tamo
908adee6fc
Fix the addition of empty payload 2025-01-09 10:24:36 +01:00
Louis Dureuil
4275833bab
Rename compute.rs to post_process.rs 2025-01-07 15:31:20 +01:00
Louis Dureuil
de7f8c4406
refactor indexer mod 2025-01-07 15:29:02 +01:00
Tamo
08fd026ebd
fix warning 2024-12-11 18:18:13 +01:00
Tamo
fa885e75b4
rename the send_progress in progress 2024-12-11 18:13:12 +01:00
Tamo
f1beb60204
make the progress use payload instead of documents 2024-12-11 18:07:45 +01:00
Tamo
867e6a8f1d
rename the send_progress field to progress since it s not sending anything 2024-12-11 16:25:01 +01:00
Tamo
6f4823fc97
make the number of document in the document tasks more incremental 2024-12-11 16:25:01 +01:00
Tamo
df9b68f8ed
inital implementation of the progress 2024-12-11 16:25:01 +01:00
Kerollmops
a751972c57
Prefer using a stable than a random hash builder 2024-12-10 14:25:53 +01:00
Kerollmops
89637bcaaf
Use bumparaw-collections in Meilisearch/milli 2024-12-10 11:52:20 +01:00
Kerollmops
f5dd8dfc3e
Rollback max memory usage changes 2024-12-09 10:26:30 +01:00
Louis Dureuil
3a11e39c01
Force max_memory to a min of 100MiB 2024-12-04 17:53:30 +01:00
Kerollmops
8ecb726683
Fix the minimun BBQueue channel threshold 2024-12-03 15:49:11 +01:00
Louis Dureuil
d040aff101 Stop allocating 1GiB for documents 2024-12-02 16:30:14 +01:00
Clément Renault
d5c07ef7b3
Manage key length conversion error correctly 2024-12-02 11:03:00 +01:00
Clément Renault
30eb0e5b5b
Rename recv and read methods to recv_action and recv_frame 2024-12-02 10:08:01 +01:00
Clément Renault
14ee7aa84c
Make sure the BBQueue is at least 50 MiB 2024-11-28 18:02:48 +01:00
Clément Renault
8a35cd1743
Adjust the BBQueue buffers to use 2% instead of 10% 2024-11-28 16:00:15 +01:00
Clément Renault
3c7ac093d3
Take the BBQueue capacity into account in the max memory 2024-11-28 15:43:14 +01:00
Clément Renault
b57dd5c58e
Remove the Vector variant and use the Vectors 2024-11-28 15:20:43 +01:00
Clément Renault
096a28656e
Fix a bug around deleting all the vectors of a doc 2024-11-28 15:15:06 +01:00
Clément Renault
cc4bd54669
Correctly construct the Embeddings struct 2024-11-28 13:53:25 +01:00
Clément Renault
58eab9a018
Send large payload through crossbeam 2024-11-28 12:01:06 +01:00
Clément Renault
e83534a430
Fix the indexer::index to correctly use the rayon::ThreadPool 2024-11-27 16:27:43 +01:00
Clément Renault
98d4a2909e
Fix the way we spawn the rayon threadpool 2024-11-27 16:05:44 +01:00
Clément Renault
cc63802115
Modify and return the IndexEmbeddings to write them later 2024-11-27 14:58:03 +01:00
Clément Renault
acec45ad7c
Send a WakeUp when writing data in the BBQueue buffers 2024-11-27 14:33:23 +01:00
Clément Renault
70802eb7c7
Fix most issues with the lifetimes 2024-11-27 14:32:42 +01:00
Clément Renault
6ac5b3b136
Finish most of the channels types 2024-11-27 14:32:26 +01:00