Repository navigation
automation (cluster): measure whether a published connector definition change materialises on every replica or only on the publishing one #27977
Description
domain:server#3 · os-elon-musk · session_016WV4czuTWNgKR22fKv7jia
Filing gate: ② an unmeasured observation that a maintainer ruling ordered measured on its own card. Ruling 6513279565 on #27428, out-of-scope item: "the dev's earlier unmeasured observation (a published connector change may materialise only on the publishing replica) is measured after this lands, on its own card". #27428 landed as PR #27941 at 2026-10-30T05:52Z. ⛔ Not a claim.
Question
#27941 added lane 4 (connector-credential.mutated) of the §6.2 cluster-invalidation family. Through it, a peer re-runs its connector reconcile when a credential is written. The open question is the definition side. When a connector instance's declared definition changes through a metadata publish on one replica, does every peer re-materialise it? The candidate paths are lane 1 (metadata.changed) or lane 2 (metadata.mutated) reaching the automation plugin's reconcile. The alternative is that only the publishing replica rebuilds, until a reload or restart.
Wanted
- A two-kernel measurement over one store and one shared bus, built like fix(service-automation): a connector credential written through the door reaches every replica #27941's
connector-credential-cluster.test.ts. Publish a changed connector definition on node A and read node B's live instance. - If B converges, pin that and close the card.
- If B does not converge, report which lane should carry the change, under the §6.2 principle of one signal and no second propagation channel (triage 6490140325). Do not build a fix without a ruling on that lane.
Reader
domain:server (service-automation, service-cluster).
Dedupe
#27428 / PR #27941 (credential lane). No open card covers definition propagation.
Activity
objectstack-fleet commentedon Oct 30, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: 审批与自动化 | none | P②
Triage: p2 ·
domain:server·area:automation· measure · 2026-10-30T06:58Z⛔ Not a claim.
6513279565on service-automation: a connector credential written through the door reaches only the replica that took it; a held instance on any other replica stays down until a metadata reload or a restart #27428, and fix(service-automation): a connector credential written through the door reaches every replica #27941 has landed. Use two replicas and one definition publish, and record whether the peer re-materialises through lane 1 or lane 2.pm:queue.Generated by Claude Code
objectstack-fleet commentedon Oct 30, 2026
domain:server#3 · os-elon-musk · session_016WV4czuTWNgKR22fKv7jia
Claim: PM loop round 2 · 2026-10-30T07:04Z
Session:
session_016WV4cx4GVsYbtK2oE4scBf(card thread; seatsession_016WV4czuTWNgKR22fKv7jia)Account:
os-elon-muskBranch:
claude/project-thread-8uxskf(the card thread's platform-assigned branch — the recorded platform exception; amended by the seat when the thread reported it)Worktree:
objectstack-issue-27977Domain:
domain:serverSeat:
domain:server#3File surface: one new two-kernel test under
packages/services/service-automation/src/(built likeconnector-credential-cluster.test.ts),.changeset/*only if the measurement lands as a pin. Measurement only: any runtime edit is ⛔ out of surface; a non-converging result is reported, and the fix card is the seat's to file.Container & model:
S,mode:thread,model: opus (TIER_DEFAULT),effort: highClause-②: no
Metadata-protocol: none
Responsibility: §6.2 cluster-invalidation behaviour for a connector definition published on one replica.
Thread-read: 6516940014
Premise: unmeasured (the card is the measurement, ordered by ruling 6513279565 on #27428)
Serial constraints cleared: #27428 landed as PR #27941 (2026-10-30T05:52Z). No open PR touches the connector-reconcile or cluster-lane files at claim time.
Ruling-ref: 6513279565
objectstack-fleet commentedon Oct 30, 2026
os-dev-report
domain:server#3 · os-elon-musk · session_016WV4cx4GVsYbtK2oE4scBf
{ "issue": 27977, "status": "done", "branch": "claude/project-thread-8uxskf", "pr": "https://github.com/objectstack-ai/objectstack/pull/28010", "session": "session_016WV4cx4GVsYbtK2oE4scBf", "premise_still_valid": true, "summary": "Measured: a connector definition published on replica A does NOT materialise on replica B. Two ObjectKernels (ObjectQLPlugin protocol + AutomationServicePlugin) over one shared SQLite file and one shared bus, lanes 2 and 4 attached as MetadataClusterBridgePlugin does. A per-item publish (draft save + publishMetaItem) changing providerConfig.endpoint re-materialises A (A's own metadata:reloaded via ObjectQLPlugin.announcePerItemPublish); lane 2 delivers connector/NAME into B's protocol mutation listeners; B's provider is never called. Control: a metadata:reloaded on B converges B from its own read of the shared store, so the gap is the trigger only. Cause read from code: service-automation's subscribeFlowMutations returns for any type but flow, and metadata:reloaded is kernel-local. The gap is pinned as it.fails citing #27977 and #28002 (the seat's fix card), with a plain-it harness-health case and a plain-it control. Recommendation only (lane is triage's call, recorded on #28002): lane 2, which already reaches the peer's local listeners, so the fix widens the existing listener to queue a connector reconcile, no second channel.", "tests": "connector-definition-cluster.test.ts at 022ad12c: Tests 2 passed | 1 expected fail (3). The it.fails case run temporarily as a plain it fails on exactly one assertion: expected undefined to be 'https://v2.billing.example.test' (B's last materialised endpoint). pnpm --filter @objectstack/service-automation test: Test Files 222 passed (222), Tests 2832 passed | 1 expected fail (2833), lock VERDICT command-exit 0. typecheck: lock VERDICT command-exit 0. No ablation: no runtime behaviour changed.", "gates": "dispatch-gates --commands --repo objectstack-ai/objectstack at 022ad12c: 50 families over 1 path. 49 exit 0; pnpm check:dual-build-cjs-loads exit 3 (PREREQUISITE NOT MET, needs every package's dist) = NOT MEASURED locally, CI runs it; check-plugin-teardown-shape --self-test exit 3 on the shallow clone, exit 0 after git fetch --unshallow. --ran: 50 derived, 49 run, 1 NOT-MEASURED, 0 UNRUN. pnpm lint exit 0. PR check-runs: in_progress at report time.", "dogfood_battery": "n/a — no `!` in the PR title and no (narrowing) arm", "mcp_calls": "6 — reads only: issue_read x2 (#27977 get, get_comments), pull_request_read x2 (#27963 get as body template, #28010 get as read-back), ToolSearch loads excluded", "api_writes": "2 — relay pr_create (#28010, label skip-changeset), relay comment (this report)", "open_questions": [], "out_of_scope_findings": [ "class: a · reach: public door PUT /api/v1/meta/connector/:name (saveMetaItem, no mode) answers success and the writing replica keeps serving the old definition (probe: A stayed on V1 after saving V2; A's mutation listeners heard connector/NAME) · same root as the measured gap · carrier: #28002 (seat recorded it there) · dedupe: connector direct save materialise, onMetadataMutation connector, subscribeFlowMutations flow only" ] }Generated by Claude Code
objectstack-fleet commentedon Oct 30, 2026
domain:server#3 · os-elon-musk · session_016WV4czuTWNgKR22fKv7jia
ACCEPT → PR #28010 comment 6517830278 · 2026-10-30T07:58Z · ready + auto-merge armed through the relay. Measured: the peer does not converge; fix card #28002.