feat(ext4): pack metadata + larger flex group on rootfs mkfs#2566
feat(ext4): pack metadata + larger flex group on rootfs mkfs#2566ValentaTomas wants to merge 1 commit intomainfrom
Conversation
Pass -G 64 (flex group size, default 16) and -E packed_meta_blocks=1 to mkfs.ext4. Both options improve spatial clustering of ext4 metadata on disk: - -G 64 packs inode tables and bitmaps for 64 block groups into one contiguous run instead of 16, so small-file metadata churn is more likely to dirty blocks within the same 4 MiB GCS chunk. - packed_meta_blocks places all metadata at the start of the disk so superblock / group descriptor / inode table / bitmap dirties cluster together rather than being interleaved with data regions. Net effect: fewer distinct 4 MiB chunks per snapshot diff, fewer GCS round-trips on cold resume. mkfs-time only; affects newly-built bases.
PR SummaryMedium Risk Overview Reviewed by Cursor Bugbot for commit 5c017fc. Bugbot is set up for automated code reviews on this repo. Configure here. |
❌ 98 Tests Failed:
View the full list of 104 ❄️ flaky test(s)
To view more test analytics, go to the Test Analytics Dashboard |
CI Feedback 🧐A test triggered by this PR failed. Here is an AI-generated analysis of the failure:
|
|
Close for now, will reopen after measuring later. |
|
Close for now, will reopen after measuring later. |
|
Close for now, will reopen after measuring later. |
Pass
-G 64and-E packed_meta_blocks=1tomkfs.ext4on rootfs creation. Both options cluster ext4 metadata more tightly on disk so small-file metadata dirties land in fewer 4 MiB GCS chunks per snapshot.mkfs-time only, no semantic change. Pairs with #2563 (inline_data).