Remove atime references to FilesystemStore #1584
Conversation
There was a problem hiding this comment.
Reviewable status: 0 of 1 LGTMs obtained, and 0 of 35 files reviewed, and pending CI: Analyze (javascript-typescript), Analyze (python), Bazel Dev / macos-13, Bazel Dev / macos-14, Bazel Dev / ubuntu-24.04, Cargo Dev / macos-13, Cargo Dev / ubuntu-22.04, Coverage, Installation / macos-13, Installation / macos-14, Installation / ubuntu-22.04, Local / lre-cc / ubuntu-24.04, Local / lre-rs / macos-14, Local / lre-rs / ubuntu-24.04, NativeLink.com Cloud / Remote Cache / macos-14, NativeLink.com Cloud / Remote Cache / ubuntu-24.04, Publish image, Publish nativelink-worker-init, Publish nativelink-worker-lre-cc, Remote / lre-cc / large-ubuntu-22.04, Remote / lre-rs / large-ubuntu-22.04, Web Platform Deployment / ubuntu-24.04, asan / ubuntu-22.04, docker-compose-compiles-nativelink (22.04), integration-tests (22.04), macos-13, pre-commit-checks, ubuntu-20.04 / stable, ubuntu-22.04, ubuntu-22.04 / stable, vale, windows-2022 / stable, and 1 discussions need to be resolved (waiting on @aaronmondal)
a discussion (no related file):
+@aaronmondal
There was a problem hiding this comment.
Reviewed 7 of 8 files at r3, 28 of 28 files at r4, all commit messages.
Reviewable status: 1 of 1 LGTMs obtained, and all files reviewed, and 3 discussions need to be resolved
-- commits line 10 at r4:
nit: duplicate "atime then atime"
nativelink-store/src/filesystem_store.rs line 480 at r4 (raw file):
.accessed() .or_else(|_| metadata.modified()) .unwrap_or(SystemTime::UNIX_EPOCH);
nit: Would this work on windows?
FilesystemStore will no longer keep atime of files up-to-date. This was an expensive operation, since every time any CAS/AC item was touched, it'd update the atime, which caused a lot of sys calls. While the system is running this PR has no effect, but when the program is restarted, items are now inserted in order of atime then mtime, but it relies on the filesystem to keep atime up-to-date. This means that .has() calls do not "refresh" the atime on disk.
There was a problem hiding this comment.
Reviewable status: 1 of 1 LGTMs obtained, and all files reviewed, and pending CI: pre-commit-checks, and 3 discussions need to be resolved
nativelink-store/src/filesystem_store.rs line 480 at r4 (raw file):
Previously, aaronmondal (Aaron Siddhartha Mondal) wrote…
nit: Would this work on windows?
Yes, it should. We are using standard library stuff here and rust is pretty good at gating linux specific things behind a different include ext.
There was a problem hiding this comment.
Reviewable status: 1 of 1 LGTMs obtained, and all files reviewed, and pending CI: pre-commit-checks, and 1 discussions need to be resolved
Previously, aaronmondal (Aaron Siddhartha Mondal) wrote…
nit: duplicate "atime then atime"
Done.
There was a problem hiding this comment.
Reviewed all commit messages.
Reviewable status: 1 of 1 LGTMs obtained, and all files reviewed, and 1 discussions need to be resolved
There was a problem hiding this comment.
Reviewable status:
complete! 1 of 1 LGTMs obtained, and all files reviewed
FilesystemStore will no longer keep atime of files up-to-date. This
was an expensive operation, since every time any CAS/AC item was
touched, it'd update the atime, which caused a lot of sys calls.
While the system is running this PR has no effect, but when the
program is restarted, items are now inserted in order of atime then
mtime, but it relies on the filesystem to keep atime up-to-date.
This means that .has() calls do not "refresh" the atime on disk.
This change is