raid1 seeded rebuild — monotonic reintegration (SPEC-74 M6, stacked on #13)#17
Open
rducom wants to merge 2 commits into
Open
raid1 seeded rebuild — monotonic reintegration (SPEC-74 M6, stacked on #13)#17rducom wants to merge 2 commits into
rducom wants to merge 2 commits into
Conversation
…ion, no pause
Design for SPEC-74 M6 (spdk-csi): the joining raid1 member enters the write
path immediately (write-only attach: write-replicated, read-excluded) and the
historical delta — FIXED by an epoch freeze taken AFTER the attach (H1 folds
the interstice) — is backfilled by the in-raid rebuild engine seeded with CBT
dirty ranges (fast-advance across clean gaps, per-window quiesce_range region
lock unchanged). Correctness argument in the doc: clean-gap skipping is safe
only BECAUSE the member is already write-replicated everywhere; the copy reads
current healthy-leg content under the window quiesce, so no external-copy
clobbering path exists.
Grounded on verified mechanics: whole-range watermark process (bdev_raid.c),
ch_processed split routing, raid1 read/write both keyed on channel presence
(the one extension needed: per-member read_excluded), CBT frozen-epoch range
export, patch-0001 skip_rebuild promotion/unwind discipline, patch-0008 range
shapes, SEC1 audit.
RPC surface (two-phase): bdev_raid_add_base_bdev {write_only} +
bdev_raid_start_seeded_rebuild {name, base_bdev, ranges[]}. RPC-CONTRACT gains
P5b; P5 (pause window) remains until the C4 bench decides — burden of proof on
the window.
…eded backfill) Implements docs/SEEDED-REBUILD-DESIGN.md (SPEC-74 M6 / P5b): add_base_bdev gains write_only (member wired into all channels for WRITES, read_excluded until completion, SB flip deferred); bdev_raid_start_seeded_rebuild starts the in-raid rebuild seeded with CBT dirty ranges — clean windows complete without copy (per-window quiesce kept, V1), dirty windows copy exactly as vanilla. Completion clears read-exclusion and flips the SB slot (0001 discipline). Failure follows the vanilla member-removal path; ranges freed on every teardown. SEC1-audited RPC.
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.
Stacked on #13. First commit is the DESIGN + P5b RPC contract (docs/SEEDED-REBUILD-DESIGN.md): the joining member attaches write-replicated/read-excluded, an epoch freeze taken AFTER the attach fixes the delta (H1 folds the interstice), and the in-raid rebuild engine backfills seeded with CBT dirty ranges — fast-advance across clean gaps, per-window quiesce_range unchanged, NO admission pause anywhere on the path.
Implementation commits follow on this branch: raid1 read_excluded decoupling, seeded raid_bdev_process, the two RPCs (add_base_bdev{write_only} + start_seeded_rebuild), tests. Gate: SPEC-74 C4 bench (p99-under-roll A/B vs today's ~4s pause window) — burden of proof on the window.