atelet: Hardlink durable-dir data out of a local checkpoint - #1514
Open
Chenyi Wang (chw120) wants to merge 1 commit into
Open
atelet: Hardlink durable-dir data out of a local checkpoint#1514Chenyi Wang (chw120) wants to merge 1 commit into
Chenyi Wang (chw120) wants to merge 1 commit into
Conversation
Staging a local checkpoint for a restore byte-copied every file the snapshot named. The durable-dir data is the largest of them -- it is the actor's own files -- and nothing after staging writes it: ateom extracts the tar into the durable-dir mount and re-archives from that mount, into a different directory, at the next checkpoint. The copy therefore spent a second full write of the actor's data on every restore and left that many dirty pages behind for the guest's first flush to queue against. Hardlink that file instead, and keep copying everything else. The distinction is the guest memory image: it is the one staged file a restore can write through, and a shared inode would carry that write back into the local checkpoint, which is the source of truth for every restore after this one. The test is an allow-list so that a snapshot file nobody has classified yet gets the safe treatment, and a failed link falls back to a copy, since the two directories sitting on different filesystems is a node's layout rather than a fault.
Chenyi Wang (chw120)
force-pushed
the
atelet-hardlink-checkpoints
branch
from
September 5, 2026 04:09
8bcfe95 to
1005853
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Staging a local checkpoint for a restore byte-copied every file the snapshot named. The durable-dir data is typically the bulk of it -- it is the actor's own files -- and nothing after staging writes it: ateom extracts the tar into the durable-dir mount and re-archives from that mount, into a different directory, at the next checkpoint. The copy therefore spent a second full write of the actor's data on every restore from a local checkpoint, and left that many dirty pages behind for the guest's first flush to queue against.
Hardlink that file instead, and keep copying everything else. The distinction is the guest memory image: it is the one staged file a restore can write through, and a shared inode would carry that write back into the local checkpoint, which is the source of truth for every restore after this one. The check is an allow-list so that a snapshot file nobody has classified yet gets the safe treatment, and a failed link falls back to a copy, since two directories on different filesystems is a node's layout rather than a fault.