Repository navigation
fix(OMN-19833): projection writers report rows_upserted and route no pure fold, and a gate keeps it so - #3100
Merged
jonahgabriel merged 3 commits intoSep 30, 2026
Conversation
…es the way the runtime expects Measured on the .201 dev lane 2026-09-30T03:27Z-03:39Z when the landing orchestrator's events first reached the bus (OMN-20127 hot-patch): the runtime went DEGRADED on projection_apply_divergence, failing every compose-dev lab pass, although the writer had landed 9 state rows and 34 transition rows in omnidash_analytics.omninode_internal. Three causes, each pinned by a test that fails on the previous code: 1. The writer reported its count as rows_written. The runtime reads only rows_upserted (omnibase_infra _extract_rows_upserted) and reads any other key as 0, so every real write counted as zero. The count now sits under rows_upserted, and a stale-write refusal under the runtime's refusal key, so a redelivery logs as a refusal, not a silent zero. 2. The pure fold was routed beside the writer. It was handed the raw event, failed nine extra_forbidden errors per message, and its zero upserts also counted toward the divergence. Only the writer is routed now, as OMN-19721 did for its siblings; the writer calls the fold in process. 3. One writer instance is routed on four topics and the runtime runs it in worker threads, so a transitioned and an agent-needed event of the same transition overlapped on the one adapter's pool (PoolConnectionHolder.wait_until_released). Each message now opens its own adapter, on its own loop, with the DSN the runtime bound. The tables were never absent: database_ref application resolves to omnidash_analytics on the dev lane, which the first diagnosis did not read. Onex-Lane: landing-projection-0510 Onex-Session: 5634e1aad3db43d594e27f74df90c7c9
…pure fold, and a gate keeps it so The runtime counts a projection write only under rows_upserted. The ci attempt outcome and worktree reconcile writers reported rows_written, so every real write counted as zero (omnimarket#3098 fixed the same for pr landing). Nine contracts also routed a pure fold beside the in-process writer; the runtime hands the fold the raw event, so it can never report a write. Each now routes the writer only, which calls the fold in process. New gate omnimarket.validators.projection_writer_contract (pre-commit hook and a pytest over the real tree plus positive controls) refuses a writer that never reports rows_upserted and a routed fold beside an in-process writer.
jonahgabriel
enabled auto-merge (squash)
September 30, 2026 07:06
|
| Rule | Description |
|---|---|
| ARCH-TOPIC-001 | No hardcoded Kafka topic strings in handler code |
| ARCH-DI-001 | No event_bus=None bypass |
| ARCH-DI-002 | No direct Handler instantiation outside workflow_runner/adapters |
| ARCH-DI-003 | No reinvented DI containers |
| ARCH-TOPIC-002 | contract.yaml topics follow `onex.{cmd |
ARCH-DI-002: Direct Handler instantiation (warning)\n\n```\nsrc/omnimarket/nodes/node_projection_ci_attempt_outcome/handlers/handler_ci_attempt_outcome_writer.py:123: self._derive = HandlerProjectionCiAttemptOutcome()
src/omnimarket/nodes/node_projection_metering_summary/handlers/handler_metering_summary_writer.py:96: result = HandlerProjectionMeteringSummary().handle(request)
src/omnimarket/nodes/node_projection_metering_summary/handlers/handler_projection_metering_summary.py:54: summary = HandlerMeteringSummary().handle(
src/omnimarket/nodes/node_projection_pr_landing/handlers/handler_pr_landing_writer.py:223: self._derive = HandlerProjectionPrLanding()
src/omnimarket/nodes/node_projection_worktree_reconcile/handlers/handler_worktree_reconcile_writer.py:49: self._derive = HandlerProjectionWorktreeReconcile()
src/omnimarket/projection/sqlite_metering_summary.py:42: result = HandlerProjectionMeteringSummary().handle(\n```\n\nHandler classes should be wired only in workflow_runner.py or adapter modules. Direct instantiation elsewhere bypasses the contract system.\n\n
Static architectural lint — no model inference (OMN-14176).
onexbot-occ-writer Bot
pushed a commit
to OmniNode-ai/onex_change_control
that referenced
this pull request
Sep 30, 2026
…mniNode-ai/omnimarket#3100 OCC companion by node_pr_lifecycle_fix_effect (OMN-13317 F1 / OMN-13990 / OMN-14285). Product PR head f0b55b603032b6e3ff160a91760cda734272cafd.
onexbot-occ-writer Bot
added a commit
to OmniNode-ai/onex_change_control
that referenced
this pull request
Sep 30, 2026
…nimarket#3100 The declared checks were executed for real in the OmniNode-ai/omnimarket checkout at the PR head by the product-repo OCC receipt runner, and the results written append-only into this companion. Producer-minted PENDING receipts are superseded by a net-new record; no existing receipt was edited. Ticket: OMN-16859 (AC3b) Onex-Workflow: OCC Receipt Runner Onex-Run: https://github.com/OmniNode-ai/omnimarket/actions/runs/36681848018
✅ Hostile Reviewer — PASSEDBlocking findings (critical/error): 0 Gate semantics
Powered by omniintelligence.review_pairing.cli_review — node-based adversarial review via HandlerLlmCliSubprocess (OMN-8524) |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| def _dispatch(event: dict[str, Any]) -> None: | ||
| try: | ||
| results.append(writer.handle(dict(event))) | ||
| except BaseException as exc: |
| pytestmark = pytest.mark.unit | ||
|
|
||
| _NODES_ROOT = Path(gate.__file__).resolve().parent.parent / "nodes" | ||
| _TOPIC = "onex.evt.test.thing-happened.v1" |
…e Evidence-Source on f0b55b6 Local readback on lab host h202: 82 passed, 11 skipped over the nine test files this PR touches. No code change.
onexbot-occ-writer Bot
added a commit
to OmniNode-ai/onex_change_control
that referenced
this pull request
Sep 30, 2026
…nimarket#3100 The declared checks were executed for real in the OmniNode-ai/omnimarket checkout at the PR head by the product-repo OCC receipt runner, and the results written append-only into this companion. Producer-minted PENDING receipts are superseded by a net-new record; no existing receipt was edited. Ticket: OMN-16859 (AC3b) Onex-Workflow: OCC Receipt Runner Onex-Run: https://github.com/OmniNode-ai/omnimarket/actions/runs/36684098258
jonahgabriel
deleted the
jonah/omn-19833-projection-writers-rows-upserted
branch
September 30, 2026 09:35
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
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.
Summary
rows_upserted. The ci attempt outcome and worktree reconcile writers reportedrows_written, so every real write counted as zero: a zero-row error per message, no terminal event, and theprojection_apply_divergencehealth dimension degraded. Both now reportrows_upserted. (Same defect omnimarket#3098 fixed for pr landing.)omnimarket.validators.projection_writer_contract: refuses an in-process writer that never reportsrows_upsertedand a routed fold beside an in-process writer. Wired as a pre-commit hook and astests/validators/test_projection_writer_contract_omn19833.py, which runs in CI over the real tree, with synthetic positive controls and a check that the key and routing agree with the runtime's own_extract_rows_upsertedand_topics_for_handler_entry. A vacuity floor fails a scan that sees fewer than 12 writers (15 today).Stacking
This branch sits on the head of omnimarket#3098 (its commit is the first in this PR). Once that merges, this rebases to the two-file-class delta above.
Audit
All 15 in-process projection writers were read. The other 13 already report
rows_upserted. Writers that are standalone runners (no in-process declaration) are out of scope by design.Test plan
_extract_rows_upsertedreads their count.Evidence-Ticket: OMN-19833
Evidence-Source: OCC#11871
Dev lane readback (hot-patched into omninode-runtime, warm restart, 2026-09-30)
rows_upserted: 1, projected: true. Before, that count read as zero.ci_attempt_outcomerow stored, applied event carriedrows_upserted: 1. No zero-row error line for either handler after the restart. Probe rows deleted afterwards.projection_apply_divergenceHEALTHY, all ten dimensions HEALTHY.