* Unify parking_lot/arc_lock feature
* Move lib/common/{io,memory}/* -> lib/common/common/*
- Mmap-related items are grouped into `common::mmap` sub-module:
- `memory/src/chunked_utils.rs` -> `common/src/mmap/chunked.rs`
- `memory/src/madvise.rs` -> `common/src/mmap/advice.rs`
- `memory/src/mmap_ops.rs` -> `common/src/mmap/ops.rs`
- `memory/src/mmap_type_readonly.rs` -> `common/src/mmap/mmap_readonly.rs`
- `memory/src/mmap_type.rs` -> `common/src/mmap/mmap_rw.rs`
- Filesystem-related items are grouped into `common::fs` sub-module:
- `common/src/fs.rs` -> `common/src/fs/sync.rs`
- `io/src/file_operations.rs` -> `common/src/fs/ops.rs`
- `io/src/move_files.rs` -> `common/src/fs/move.rs`
- `io/src/safe_delete.rs` -> `common/src/fs/safe_delete.rs`
- `memory/src/checkfs.rs` -> `common/src/fs/check.rs`
- `memory/src/fadvise.rs` -> `common/src/fs/fadvise.rs`
- Rest is moved straight into `common`:
- `io/src/storage_version.rs` -> `common/src/storage_version.rs`
The old `io` and `memory` are now hollow crates that re-export items
from `common`. These hollow crates will be removed in next commits.
* Replace uses of `io` and `memory` with new paths in `common`
Since `io` and `memory` are just re-exports of `common`, these
replacements are no-op.
* Remove `io` and `memory` crates
* weighted rrf implementation
* test
* fmt
* fix edge
* validate number of sources and number of weights
* do not partial match
* upd schema
* review fixes
* update formula
* remove calcualtions from tests
* update comment, because AI have OCD
* fmt
* download tar
* compute sha256 for stream download
* wip: propagate unpacking into down to the logic, todo: validation
* unpacked snapshot validation
* Minor tweaks
* Fix typo
* validation during unpack
* cancellation token
* update docstring
* remove redundant dep
* Rearrange unpack functions
- Rename `safe_unpack.rs` into `tar_unpack.rs` so it would be listed
near `tar_ext.rs` in IDEs.
- Replace calls like `ar = open_snapshot_archive(…); safe_unpack(ar, …);`
with a single call to `tar_unpack_file(…)`.
- Put calls to `Archive::new(); Archive::set_overwrite(false);` inside
`tar_unpack_reader` (was `safe_unpack`). So, now it is the only place
that does `set_overwrite`.
* Let clippy complain if tar::Archive::unpack used
* Mock snapshot download URL
Instead of downloading from storage.googleapis.com every time the test
runs, put small snapshot file to the repo.
The snapshot file is created using this command:
curl -s \
https://storage.googleapis.com/qdrant-benchmark-snapshots/test-shard.snapshot \
| tar \
--delete segments/4ea958d8-0b64-4312-9a53-0cd857e93535.tar \
--delete segments/65ac6276-8cca-4f5c-b767-9722190cee8b.tar \
> lib/storage/src/content_manager/snapshots/test-shard.snapshot
File contents:
$ tar tf lib/storage/src/content_manager/snapshots/test-shard.snapshot
wal/
wal/closed-255
newest_clocks.json
replica_state.json
shard_config.json
$ du -sh lib/storage/src/content_manager/snapshots/test-shard.snapshot
12K lib/storage/src/content_manager/snapshots/test-shard.snapshot
$ sha256sum < lib/storage/src/content_manager/snapshots/test-shard.snapshot
5d94eac5c1ede3994a28bc406120046c37370d5d45b489a0d2252531b4e3e1f2 -
---------
Co-authored-by: timvisee <tim@visee.me>
Co-authored-by: xzfc <xzfcpw@gmail.com>
* Remove call to Archive::set_sync
Also, this method was the last remaining part of our `tar-rs` fork,
so we can switch to the upstream version now.
* Do syncfs
* Use RwLock for pending changes in MmapBitSliceBufferedUpdateWrapper
* Use RwLock for pending operations in DatabaseColumnScheduledDeleteWrapper
* Use RwLock for pending updates in DatabaseColumnScheduledUpdateWrapper
* Drop alive guard before reconciliation, we don't touch files after
* Remove redundant clone
* Update comments
* in MmapSliceBufferedUpdateWrapper
* in MmapBitsliceBufferedUpdateWrapper
* in MutableIdTracker's versions updates
* in MutableIdTracker's mapping updates
* clone updates only when non-empty
* only lock for reconciling pending changes
* simpler reconciling
* use Mutex as argument to ensure we only lock within reconciliation
* introduce IsAliveLock
* add tests
* rename to mark_dead
* update gridstore comment
* Update some comments
* Add must_use attribute to lock_if_alive
* Rename mark_dead to blocking_mark_dead
* make the handle take a `Weak` reference
* make dropping explicit
* use `Mutex::lock_arc` instead
* less nesting
* clippy
---------
Co-authored-by: timvisee <tim@visee.me>
* Initial plan
* Update timeout error handling to use milliseconds for better precision
Co-authored-by: agourlay <606963+agourlay@users.noreply.github.com>
* Use Duration type and debug format for timeout errors
Changed timeout() methods to accept Duration directly instead of u128 milliseconds.
This uses Rust's built-in Duration debug formatting ({:?}) which provides clean output:
- Sub-second: "500ms"
- Exact seconds: "1s"
- Mixed: "2.5s"
Updated all call sites to pass Duration directly instead of timeout.as_millis().
Co-authored-by: agourlay <606963+agourlay@users.noreply.github.com>
* fmt/clippy
* make it consistent
* Inline some arguments and imports
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agourlay <606963+agourlay@users.noreply.github.com>
Co-authored-by: Arnaud Gourlay <arnaud.gourlay@gmail.com>
Co-authored-by: timvisee <tim@visee.me>
* Timeout on point retrieve when acquiring segment holder
* review: less radical and cleaner
* highly embarassing I hope no one sees this!
* coderabbit
* no manual impl for Serialize and Hash
* new feedback query
* construct query scorer
* simplify formula
* better comment
* make it easier to add other feedback formulas
* rename query -> target
* rename in async raw scorer
* rename formula -> strategy
* better comments
* rename to Simple strategy
* clippyyyyy
* oops, reenable conditional compilation
* Fundamental refactor to precompute pair score
* rename FeedbackScorer -> FeedbackQuery
This type implements `Query` trait, not `Scorer`.`
* wip: generalization trait for queries
* implement generalization for point operations
* fmt
* log priority queue
* wip: SlowRequestsListener
* fmt
* fix clippy
* simplify generalization
* fmt
* implement collection of requests profiles for update API
* implement API for viewing slow requests log
* add collection name to update worker
* add datetime to log
* fmt
* probabilistic counter of unique requests
* rename
* compute hash before converting into json value
* move logable out of generalizable
* fmt
* log query request
* fmt
* some fixes
* move measurement into local shard
* fmt
* upd openapi (not important)
* For enum variants, has discriminant
* Make SearchParams Copy
* Hash 0.0 and -0.0 the same
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* Correctly hash enum variants and float values
* Hash through ordered float instead
* Fix priority queue not keeping longest request for hash
* SearchParams implements Copy
* Fix clippy warning
* Add unordered_hash_unique
* skip serialization if none
* Use OrderedFloat for hashing a float
* Use OrderedFloat for hashing a float
* only log updates if they are performed
---------
Co-authored-by: timvisee <tim@visee.me>
Co-authored-by: Tim Visée <tim+github@visee.me>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Co-authored-by: xzfc <xzfcpw@gmail.com>
* allow custom K parameter for RRF
* generate grpc docs and openapi
* use tagged type approach for parametrized fusions
* use params approach in grpc
* simplify api structure
* upd schema
* nits
* rest: parameterized rrf as query variant
* consistency in doc comments
---------
Co-authored-by: Andrey Vasnetsov <andrey@vasnetsov.com>
* put all flags impls in a single module
* use in appendable_dense_vector_storage
* use in appendable_mmap_multi_dense_storage
* use in mmap_sparse_vector_storage
* clippy
* Add comment we cannot rely on length of deleted vectors bitvec
---------
Co-authored-by: timvisee <tim@visee.me>
* separate flags buffering and in-memory container
* new bitvec flags
* fmt
* clippy
* make BufferedDynamicFlags public to crate only
* clear cache after loading
* clippyyy
* improve comment
* Use resize with a constant to eliminate a hot closure
* Add comment on why we take subslice of bitslice
---------
Co-authored-by: timvisee <tim@visee.me>
* Introduce BufferedDynamicFlags
* clippy and extra changes
* fix flushing
* add mutable index impl
* use MutableNullIndex instead of MmapNullIndex
* use BufferedDynamicFlags
* remove mmap_null_index
* take updates before returning the flusher
* bump total_point_count on add and remove point
* improve test case
* Only clean up mutable null index files if index exists
* Remove mention of cached in context of length
* refactor flags into a single `Option<Storage>`
---------
Co-authored-by: timvisee <tim@visee.me>
* Add function to detect if storage is using RocksDB
* Add feature flag to toggle payload index migration
* Add naive payload index migration, rebuild from scratch if RocksDB
* Adjust payload index rebuild logs
* When rebuilding payload indices, clean up existing storages
* Update OpenAPI spec
* Use weak references to Gridstore in mutable payload index flushers
* Rework payload index clear/wipe functions, make them consistent
* Naively migrate RocksDB before loading indices
* Fix mmap payload index load, staged like others (#6837)
* Add staged loading of mmap numeric payload index
* Add staged loading of mmap map payload index
* Add staged loading of mmap geo payload index
* Add staged loading of mmap full text payload index
* Add staged loading of mmap bool and null payload index
* Minor tweaks
* Don't fail clearing file cache if file does not exist
---------
Co-authored-by: jojii <jojii@gmx.net>
* Fix not changing storage states, migrate vector storages in full segment
* Destroy old vector storage data after updating segment config
* Disable existing migrations, only use segment level ones
* Remove segment config update warnings
* Correctly bump sparse vector storage type
* Fix clippy warnings without RocksDB feature flag
* Flag index selector
* Flag take_database_snapshot
* Flag field index builder
* Flag bool index
* Flag full text index
* Flag geo index
* Flag map index
* Feature flag numeric index
* Flag skip_rocksdb
* Flag payload storage
* Flag segment
* Flag builder
* Flag backup
* Make RocksDB crate optional
* Tweak feature flag gate
* Disable default segment crate features from all dependents
* Handle flag in collection crate correctly
* Flag payload storage types in tests
* Temporarily disable on-disk check for sparse vectors
* Flag sparse vector RocksDB in GPU code
* Fix compilation error since recent merge
* Flag payload storage types in new function
* remove todo
* Explicitly implement default to more clearly state difference
---------
Co-authored-by: generall <andrey@vasnetsov.com>
* Actively migrate RocksDB ID tracker to new format on segment load
* Extract ID tracker migration logic to function
* Feature flag ID tracker migration
* Simplify ID tracker migration by moving it deeper into load function
* Move migrate function to the bottom
* Add test to assert RocksDB to mutable ID tracker migration
* Assert new mutable ID tracker is empty
* Review remarks
* On RocksDB to mutable ID tracker migration failure, clean up files
* Demote empty mutable ID tracker to debug assertion
* Copy all point versions, including deleted, set known mappings
* read links and versions separatelly
---------
Co-authored-by: generall <andrey@vasnetsov.com>
* Improve telemetry logic and test
* Parametrize telemetry test
* Consistency hash peeer ID across telemetry
* clean test
* Use Option in segments telemetry
* updat openapi spec
* Avoid test failure on change in order of params
* Add bench for mmap_bitslice_buffered_update_wrapper
* Use ahash in buffered mmap/RocksDB types and in payload storage
* Function is used in testing only
* fix flush routine for persistent wrappers
* fmt
* rollback changes for write-only wrappers
* Use retain to drop flushed items
* Rename function
---------
Co-authored-by: timvisee <tim@visee.me>
* Add in rest and grpc
* add to QueryEnum
* implement Query trait
* connect to scorer creation
* upd tests
* additional changes
* fmt
* gen openapi and grpc docs
* coderabbit fix
* add changes in async scorer
* test sum_scores in more places, refactor to remove repetition
* Handle missing CF on the payload index flush level
* move error handling on the payload field level, so other fields are not skipped if one fails to flush
* Merge nested match statement
---------
Co-authored-by: timvisee <tim@visee.me>
xzfc