Clear disk cache after read by xzfc · Pull Request #6396 · qdrant/qdrant

xzfc · 2025-04-16T17:44:20Z

Related: #6323.

Consider this simplified example:

fn read_data(path: &Path) -> io::Result<Vec<u8>> {
    let file = File::open(path)?;
    let reader = BufReader::new(file);
    let data = Vec::new();
    reader.read_to_end(&mut data)?;
    data
}

The problem with this code is that after calling this function, the file contents is stored twice in the memory: once as Vec<u8> (as anonymous memory), and once implicitly as OS file cache, making it more likely to evict other files from the cache.

Particularly, after reproducing #6323, I've found that quantized vectors and id tracker files are still in the OS file cache when the mmaped index cache gets evicted. Since we are unlikely to read quantized vectors again, we'd better to avoid polluting the OS file cache.

This PR adds a new wrapper around File, OneshotFile, that calls posix_fadvise(…, POSIX_FADV_DONTNEED) when the file is closed. Not ideal, but should improve the situation a bit.

coderabbitai · 2025-04-16T17:47:20Z

📝 Walkthrough

Walkthrough

This set of changes introduces a new fadvise module in the lib/common/memory crate that provides functionality for managing file access patterns and explicit cache eviction using the posix_fadvise system call on supported platforms. A new build script (build.rs) is added to conditionally enable compilation of posix_fadvise-dependent code based on the target OS and environment. The core addition is the OneshotFile struct, a wrapper around std::fs::File that applies sequential and no-reuse access hints and supports explicit cache dropping via drop_cache(). The existing clear_disk_cache function is removed from the madvise module and reimplemented in the new fadvise module. All imports and usages of clear_disk_cache across the codebase are updated to reference the fadvise module. Additionally, several components now use OneshotFile for file reading and explicitly drop file caches after operations, replacing previous uses of standard File.

Suggested reviewers

generall
timvisee

Tip

⚡💬 Agentic Chat (Pro Plan, General Availability)

We're introducing multi-step agentic chat in review comments and issue comments, within and outside of PR's. This feature enhances review and issue discussions with the CodeRabbit agentic chat by enabling advanced interactions, including the ability to create pull requests directly from comments and add commits to existing pull requests.

📜 Recent review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 51e227b and 6414477.

📒 Files selected for processing (3)

lib/quantization/src/encoded_storage.rs (2 hunks)
lib/segment/src/id_tracker/mutable_id_tracker.rs (2 hunks)
lib/segment/src/vector_storage/chunked_vectors.rs (3 hunks)

🚧 Files skipped from review as they are similar to previous changes (3)

lib/segment/src/vector_storage/chunked_vectors.rs
lib/quantization/src/encoded_storage.rs
lib/segment/src/id_tracker/mutable_id_tracker.rs

⏰ Context from checks skipped due to timeout of 90000ms (13)

GitHub Check: test-snapshot-operations-s3-minio
GitHub Check: test-shard-snapshot-api-s3-minio
GitHub Check: test-low-resources
GitHub Check: test-consistency
GitHub Check: integration-tests-consensus
GitHub Check: Basic TLS/HTTPS tests
GitHub Check: rust-tests (macos-latest)
GitHub Check: integration-tests
GitHub Check: lint
GitHub Check: rust-tests (windows-latest)
GitHub Check: test-consensus-compose
GitHub Check: rust-tests (ubuntu-latest)
GitHub Check: storage-compat-test

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

Review comments: Directly reply to a review comment made by CodeRabbit. Example:
- I pushed a fix in commit <commit_id>, please review it.
- Generate unit testing code for this file.
- Open a follow-up GitHub issue for this discussion.
Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
- @coderabbitai generate unit testing code for this file.
- @coderabbitai modularize this function.
PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
- @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
- @coderabbitai read src/utils.ts and generate unit testing code.
- @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
- @coderabbitai help me debug CodeRabbit configuration file.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

@coderabbitai pause to pause the reviews on a PR.
@coderabbitai resume to resume the paused reviews.
@coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
@coderabbitai full review to do a full review from scratch and review all the files again.
@coderabbitai summary to regenerate the summary of the PR.
@coderabbitai generate docstrings to generate docstrings for this PR.
@coderabbitai resolve resolve all the CodeRabbit review comments.
@coderabbitai configuration to show the current CodeRabbit configuration for the repository.
@coderabbitai help to get help.

Other keywords and placeholders

Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
Add @coderabbitai anywhere in the PR title to generate the title automatically.

CodeRabbit Configuration File (`.coderabbit.yaml`)

You can programmatically configure CodeRabbit by adding a .coderabbit.yaml file to the root of your repository.
Please see the configuration documentation for more information.
If your editor has YAML language server enabled, you can add the path at the top of this file to enable auto-completion and validation: # yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json

Documentation and Community

Visit our Documentation for detailed information on how to use CodeRabbit.
Join our Discord Community to get help, request features, and share feedback.
Follow us on X/Twitter for updates and announcements.

coderabbitai

Actionable comments posted: 0

🧹 Nitpick comments (1)

lib/segment/src/id_tracker/mutable_id_tracker.rs (1)

659-675: Consider using OneshotFile for loading versions as well.

The load_versions function still uses a standard File while the load_mappings function was updated to use OneshotFile. For consistency and to fully implement the cache optimization strategy, consider updating this function as well.

-    let file = File::open(versions_path)?;
+    let file = OneshotFile::open(versions_path)?;

     let file_len = file.metadata()?.len();
     if file_len % VERSION_ELEMENT_SIZE != 0 {
         log::warn!(
             "Corrupted ID tracker versions storage, file size not a multiple of a version, assuming automatic recovery by WAL"
         );
     }
     let version_count = file_len / VERSION_ELEMENT_SIZE;

     let mut reader = BufReader::new(file);

-    Ok((0..version_count)
+    let result = Ok((0..version_count)
         .map(|_| reader.read_u64::<FileEndianess>())
-        .collect::<Result<_, _>>()?)
+        .collect::<Result<_, _>>()?);
+    
+    // Drop the cache after reading
+    let file = reader.into_inner();
+    file.drop_cache()?;
+    
+    result

📜 Review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 751fb5a and e5c5748.

⛔ Files ignored due to path filters (1)

Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (23)

lib/common/memory/Cargo.toml (2 hunks)
lib/common/memory/build.rs (1 hunks)
lib/common/memory/src/fadvise.rs (1 hunks)
lib/common/memory/src/lib.rs (1 hunks)
lib/common/memory/src/madvise.rs (0 hunks)
lib/gridstore/src/bitmask/gaps.rs (1 hunks)
lib/gridstore/src/bitmask/mod.rs (1 hunks)
lib/gridstore/src/page.rs (1 hunks)
lib/quantization/src/encoded_storage.rs (2 hunks)
lib/segment/src/id_tracker/mutable_id_tracker.rs (3 hunks)
lib/segment/src/index/field_index/full_text_index/mmap_inverted_index/mod.rs (1 hunks)
lib/segment/src/index/field_index/geo_index/mmap_geo_index.rs (1 hunks)
lib/segment/src/index/field_index/map_index/mmap_map_index.rs (1 hunks)
lib/segment/src/index/field_index/mmap_point_to_values.rs (1 hunks)
lib/segment/src/index/field_index/numeric_index/mmap_numeric_index.rs (1 hunks)
lib/segment/src/index/hnsw_index/hnsw.rs (1 hunks)
lib/segment/src/vector_storage/chunked_mmap_vectors.rs (1 hunks)
lib/segment/src/vector_storage/chunked_vectors.rs (3 hunks)
lib/segment/src/vector_storage/dense/dynamic_mmap_flags.rs (1 hunks)
lib/segment/src/vector_storage/dense/memmap_dense_vector_storage.rs (1 hunks)
lib/segment/src/vector_storage/quantized/quantized_vectors.rs (1 hunks)
lib/sparse/src/index/inverted_index/inverted_index_compressed_mmap.rs (1 hunks)
lib/sparse/src/index/inverted_index/inverted_index_mmap.rs (1 hunks)

💤 Files with no reviewable changes (1)

lib/common/memory/src/madvise.rs

🧰 Additional context used

🧬 Code Graph Analysis (15)

lib/gridstore/src/bitmask/mod.rs (2)

lib/segment/src/vector_storage/quantized/quantized_vectors.rs (1)

lib/segment/src/vector_storage/dense/memmap_dense_vector_storage.rs (1)

lib/gridstore/src/page.rs (2)

lib/segment/src/index/hnsw_index/hnsw.rs (1)

lib/segment/src/index/field_index/mmap_point_to_values.rs (2)

lib/segment/src/vector_storage/chunked_mmap_vectors.rs (2)

lib/sparse/src/index/inverted_index/inverted_index_compressed_mmap.rs (2)

lib/segment/src/index/field_index/numeric_index/mmap_numeric_index.rs (2)

lib/segment/src/index/field_index/map_index/mmap_map_index.rs (2)

lib/segment/src/index/field_index/geo_index/mmap_geo_index.rs (2)

lib/quantization/src/encoded_storage.rs (1)

lib/gridstore/src/bitmask/gaps.rs (2)

lib/common/memory/src/lib.rs (1)

lib/segment/src/id_tracker/mutable_id_tracker.rs (1)

⏰ Context from checks skipped due to timeout of 90000ms (13)

GitHub Check: Basic TLS/HTTPS tests
GitHub Check: test-snapshot-operations-s3-minio
GitHub Check: test-shard-snapshot-api-s3-minio
GitHub Check: test-low-resources
GitHub Check: test-consistency
GitHub Check: rust-tests (macos-latest)
GitHub Check: test-consensus-compose
GitHub Check: rust-tests (windows-latest)
GitHub Check: integration-tests-consensus
GitHub Check: rust-tests (ubuntu-latest)
GitHub Check: integration-tests
GitHub Check: lint
GitHub Check: storage-compat-test

🔇 Additional comments (32)

Copilot

Pull Request Overview

This PR addresses the file cache pollution issue by replacing the previous disk cache clearing mechanism with a new fadvise‐based implementation using an OneshotFile wrapper. Key changes include:

Replacing all usages of memory::madvise::clear_disk_cache with memory::fadvise::clear_disk_cache.
Updating file read operations to use OneshotFile::open and invoking drop_cache() to clear the OS file cache after reading.
Adding a new fadvise module along with necessary build script and dependency adjustments.

Reviewed Changes

Copilot reviewed 24 out of 24 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
lib/segment/src/vector_storage/dense/memmap_dense_vector_storage.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/vector_storage/dense/dynamic_mmap_flags.rs	Updated ordering of import for clear_disk_cache to fadvise.
lib/segment/src/vector_storage/chunked_vectors.rs	Replaced File::open with OneshotFile::open and added drop_cache call.
lib/segment/src/vector_storage/chunked_mmap_vectors.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/hnsw_index/hnsw.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/field_index/numeric_index/mmap_numeric_index.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/field_index/mmap_point_to_values.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/field_index/map_index/mmap_map_index.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/field_index/geo_index/mmap_geo_index.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/index/field_index/full_text_index/mmap_inverted_index/mod.rs	Updated clear_disk_cache import to use fadvise module.
lib/segment/src/id_tracker/mutable_id_tracker.rs	Replaced File::open with OneshotFile::open and added drop_cache call.
lib/quantization/src/encoded_storage.rs	Replaced File::open with OneshotFile::open and added drop_cache call.
lib/gridstore/src/page.rs	Updated clear_disk_cache import to use fadvise module.
lib/gridstore/src/bitmask/mod.rs	Updated clear_disk_cache import to use fadvise module.
lib/gridstore/src/bitmask/gaps.rs	Updated clear_disk_cache import to use fadvise module.
lib/common/memory/src/madvise.rs	Removed clear_disk_cache implementation.
lib/common/memory/src/lib.rs	Added export for new fadvise module.
lib/common/memory/src/fadvise.rs	New module that implements clear_disk_cache and the OneshotFile wrapper.
lib/common/memory/build.rs	New build script to enable posix_fadvise support conditionally.
lib/common/memory/Cargo.toml	Added new dependency on delegate.

* Move clear_disk_cache to memory::fadvise * Add memory::fadvise::OneshotFile * Use OneshotFile

xzfc added 2 commits April 16, 2025 17:08

Move clear_disk_cache to memory::fadvise

41576d8

Add memory::fadvise::OneshotFile

f51fc01

xzfc requested review from generall and timvisee April 16, 2025 17:44

coderabbitai bot reviewed Apr 16, 2025

View reviewed changes

xzfc force-pushed the fadvise-after-read branch from e5c5748 to 51e227b Compare April 16, 2025 17:58

Use OneshotFile

6414477

xzfc force-pushed the fadvise-after-read branch from 51e227b to 6414477 Compare April 16, 2025 18:05

github-actions bot mentioned this pull request Apr 16, 2025

Flaky test multivector_filtrable_hnsw_test::test_multi_filterable_hnsw::case_5_recobestscore_eq #6318

Closed

xzfc requested a review from Copilot April 16, 2025 23:10

Copilot AI reviewed Apr 16, 2025

View reviewed changes

generall approved these changes Apr 17, 2025

View reviewed changes

generall merged commit 9fedc65 into dev Apr 17, 2025
17 checks passed

generall deleted the fadvise-after-read branch April 17, 2025 08:57

pull bot pushed a commit to kp-forks/qdrant that referenced this pull request Apr 21, 2025

Clear disk cache after read (qdrant#6396)

1f805b6

* Move clear_disk_cache to memory::fadvise * Add memory::fadvise::OneshotFile * Use OneshotFile

coderabbitai bot mentioned this pull request Jul 14, 2025

Fix mmap payload index load, load in stages #6860

Closed

9 tasks

coderabbitai bot mentioned this pull request Sep 29, 2025

Use fs-err #7319

Merged

coderabbitai bot mentioned this pull request Feb 24, 2026

use UniversalRead in ImmutableDenseVectors #8210

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Clear disk cache after read#6396

Clear disk cache after read#6396
generall merged 3 commits intodevfrom
fadvise-after-read

xzfc commented Apr 16, 2025

Uh oh!

coderabbitai bot commented Apr 16, 2025 •

edited

Loading

Walkthrough

Suggested reviewers

Chat

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

CodeRabbit Configuration File (`.coderabbit.yaml`)

Documentation and Community

Uh oh!

coderabbitai bot left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

xzfc commented Apr 16, 2025

Uh oh!

coderabbitai bot commented Apr 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Suggested reviewers

Chat

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

CodeRabbit Configuration File (.coderabbit.yaml)

Documentation and Community

Uh oh!

coderabbitai bot left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

coderabbitai bot commented Apr 16, 2025 •

edited

Loading

CodeRabbit Configuration File (`.coderabbit.yaml`)