Skip to content

feat: build shared cell-group multi-Raft runtime - #160

Open
forhappy wants to merge 713 commits into
mainfrom
codex/shared-cell-groups
Open

feat: build shared cell-group multi-Raft runtime#160
forhappy wants to merge 713 commits into
mainfrom
codex/shared-cell-groups

Conversation

@forhappy

@forhappy forhappy commented Oct 29, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Cellar SQL cells share RF3 Raft groups while retaining independent SQLite state, receipts, versions and effects.

This update fixes command admission during concurrent apply: commands wait for committed entries before entering SQL, keep their original deadline and obtain fresh execution authority. It fixes child receipt replay during split/merge without repeating callbacks or relaxing quorum and fencing checks. The PR also includes node-owned automatic partition scheduling and durable recovery of selected work.

Validation: 187 native execution tests pass (two process-fixture helpers ignored). Ten native transition cases and four real RustFS cases pass in both image modes. Strict all-target Clippy and repository checks pass. Earlier failing runs are retained with their source revisions. RustFS uses host-native TLS voters; independent failure domains and sustained load remain open.

CG-11 remains open: safe cancellation/recovery of staged work and competing selections, adaptive whole-group placement, configured startup, reserve and bounded fanout. The full CG-12–16 lifecycle and release gates also remain open.

Includes latest origin/main (981a5417). See progress and remaining priorities.

@forhappy forhappy changed the title docs: select shared cell groups as the Raft replacement docs: add executable cell-group multi-Raft delivery plan Oct 29, 2026
@forhappy forhappy changed the title docs: add executable cell-group multi-Raft delivery plan feat: add cell-group identities and version contracts Oct 29, 2026
@forhappy forhappy changed the title feat: add cell-group identities and version contracts feat: define cell-group multi-Raft delivery and wire contracts Oct 29, 2026
@forhappy forhappy changed the title feat: define cell-group multi-Raft delivery and wire contracts feat: build cell-group contracts and manifest storage Oct 29, 2026
@forhappy forhappy changed the title feat: build cell-group contracts and manifest storage feat: persist cell-group manifests through Raft Engine Oct 29, 2026
@forhappy forhappy changed the title feat: persist cell-group manifests through Raft Engine feat: persist and incrementally prepare cell-group manifests Oct 29, 2026
@forhappy forhappy changed the title feat: persist and incrementally prepare cell-group manifests feat: persist cell-group manifests with retention and reclamation Oct 29, 2026
@forhappy forhappy changed the title feat: persist cell-group manifests with retention and reclamation feat: build cell-group durability and enrollment Oct 29, 2026
@forhappy forhappy changed the title feat: build cell-group durability and enrollment feat: build cell-group durability, enrollment and range directory Oct 29, 2026
@forhappy forhappy changed the title feat: build cell-group durability, enrollment and range directory feat: build cell-group durability, routing and lazy birth metadata Oct 29, 2026
@forhappy forhappy changed the title feat: build cell-group durability, routing and lazy birth metadata feat: build cell-group storage and routing foundations Oct 29, 2026
@forhappy forhappy changed the title feat: build cell-group storage and routing foundations feat: add physical cell-group bootstrap and storage foundations Oct 29, 2026
@forhappy forhappy changed the title feat: add physical cell-group bootstrap and storage foundations feat: deliver cell-group multi-Raft plan and foundations Oct 29, 2026
@forhappy forhappy changed the title feat: deliver cell-group multi-Raft plan and foundations feat: add cell-group routing and durable enrollment foundations Oct 29, 2026
@forhappy forhappy changed the title feat: add cell-group routing and durable enrollment foundations feat: add cell-group consensus and durable state foundations Oct 30, 2026
@forhappy
forhappy marked this pull request as ready for review October 30, 2026 00:50
@forhappy forhappy changed the title feat: add cell-group consensus and durable state foundations feat: add shared cell-group runtime foundations Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group runtime foundations feat: add shared cell-group consensus and SQL foundations Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group consensus and SQL foundations feat: add shared cell-group consensus and bounded SQL workers Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group consensus and bounded SQL workers feat: add shared cell-group consensus and SQL recovery Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group consensus and SQL recovery feat: add shared cell-group consensus and registered execution Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group consensus and registered execution feat: add shared cell-group consensus, execution and transport Oct 30, 2026
@forhappy forhappy changed the title feat: add shared cell-group consensus, execution and transport feat: build shared cell-group multi-Raft runtime Oct 31, 2026
@forhappy
forhappy marked this pull request as draft November 3, 2026 07:45
forhappy added a commit that referenced this pull request Nov 3, 2026
@forhappy
feat(raft)!: advance shared-group cutover
f50b99a 
BREAKING CHANGE: canonical SDK and fleet consumers use shared physical
cell groups with independent logical cell histories. Retired per-cell
APIs have no compatibility path.

Publish the cutover checkpoint and its MVP acceptance scope. Public
consumer ports, managed S3 activation and complete build, lint and backend
gates remain unfinished; keep PR #160 in draft.

Refs #160
forhappy added a commit that referenced this pull request Nov 3, 2026
@forhappy
feat(raft)!: route shared cell queries
fc4a8d3 
Keep replica attempts in one invocation, original deadline and attempt
bank. Physical SQL and transport retain admission until joined; leader
advice never replaces destination authorization.

Normalize the exact SQL deadline envelope so timer races cannot change
fallback behavior. Drain example Node producers before sampling the final
Raft quiescence report, preserving its existing success predicate.

BREAKING CHANGE: shared query replies carry authenticated leader advice
at marker 1. No predecessor wire reader is retained.

Refs #160
forhappy added a commit that referenced this pull request Nov 3, 2026
@forhappy
feat(raft)!: port counter to shared groups
41ab8b4 
Use explicit namespace/archive provisioning and replicated enrollment.
Keep independent SQLite histories and receipts in one group per voter.
Preserve native/Axum read, token, retry and shutdown behavior.

Add the original public SDK shared-group gate with retained process
proofs. Other managed, ingress/admin and feature-off gates remain open.

BREAKING CHANGE: the counter uses logical histories and cell tokens.
Predecessor per-cell layouts are refused, without migration or fallback.
Replies distinguish cell sequence from physical commit witnesses.

Refs #160
forhappy added a commit that referenced this pull request Nov 3, 2026
@forhappy
chore: merge origin/main into shared groups
9df1389 
Resolve the host conflict at the shared configuration surface. Adapt
upstream bulk-seed callbacks to logical cell history and retain the
complete metrics inventory assertions.

Preserve upstream replica, snapshot, shutdown and qualification changes.
Their applicability to the shared membership owner requires its own
qualification; a legacy replica fix is not proof for that path.

Refs #160
forhappy added a commit that referenced this pull request Nov 4, 2026
@forhappy
feat(raft): own recovered node startup
f644d9d 
Resume exact seeded histories through an explicit fencing authority and a joined activation worker. Keep admission pending until real Raft activation applies, preserving original receipts and catalog intent.

Join public control services before infrastructure teardown on late startup refusal. Cover disk loss, retained restart, S3 minima and held-worker ownership.

Refs #160
forhappy added a commit that referenced this pull request Nov 4, 2026
@forhappy
fix(raft): bound warm archive reuse
d268cf0 
Warm cached destinations bypassed buffered source validation.
Check source bytes before reuse, and reuse completed file/directory
barriers only within their open owner. Keep verifier files charged
under existing ceilings to avoid rebuilding every archive image.

Refs #160
forhappy added 30 commits November 7, 2026 20:53
@forhappy
fix(raft): replace fences after repeated removal
e9b1c6d
@forhappy
fix(raft): verify collected birth member prefixes
64fbdd2
@forhappy
fix(raft): close collected birth debt at admission
6a7c473
@forhappy
fix(raft): release collected member disk debt
5b96c2d
@forhappy
docs(raft): explain removed member collection
5bc999c
@forhappy
docs(raft): record member reuse and open gates
9a324e9
@forhappy
feat(raft): resume durable member cleanup
5e03aef
@forhappy
fix(raft): retain cleanup deadline independently
8d5553c
@forhappy
docs(raft): document durable cleanup ownership
db8ef1d
@forhappy
test(raft): observe cleanup through a fresh client
be864b9
@forhappy
test(raft): isolate cleanup refusal identity
cf2699b
@forhappy
docs(raft): record durable cleanup delivery
7c16a08
@forhappy
feat(raft): resume cleanup on qualified boots
deac305
@forhappy
docs(raft): describe retained cleanup startup
ed53c03
@forhappy
fix(raft): isolate background cleanup catalog turns
6eacd0e
@forhappy
test(raft): surface retained cleanup worker failures
fed026a
@forhappy
fix(raft): park removed members on retained startup
fda3db0
@forhappy
test(raft): isolate topology controller clients
a352f77
@forhappy
docs(raft): record retained cleanup delivery
4a4e4a4
@forhappy
feat(raft): persist partition admission work
42f961e
@forhappy
feat(raft): recover partition work across coordinators
c4ed676
@forhappy
fix(raft): resolve historical partition settlement first
f03916d
@forhappy
test(raft): assert durable birth admission replay
2bd87ac
@forhappy
docs(raft): record durable partition recovery
60ff096
@forhappy
feat(raft): own automatic partition scheduling
67f6727
@forhappy
docs(raft): scope automatic partition delivery
3d08b52
@forhappy
fix(raft): wait for committed apply before SQL
6ae2c3f 
A sampled ready leader can commit another cell before execution admission.
Wait for its durable apply prefix under the original request deadline,
then obtain fresh authority. Retry only before SQL has entered; keep
routing budgets, leader fences and exact producer reconciliation intact.
@forhappy
test(raft): preserve bounded qualification waits
7e75a45 
Treat ordinary Raft packet overload as a deferred delivery under the
existing pump deadline. Wait for actual reconciliation proofs after
leader change. Deliver a generated heartbeat before capturing its
100-ms source proof, then stop peer traffic through expiry checks.
Keep final state assertions, freshness and resource limits unchanged.
@forhappy
test(raft): pace snapshot retry clock
032a896
@forhappy
docs(raft): record command readiness delivery
1f2d895
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@forhappy