[gaspergrom]

Summary

• Deletes pipes/health_score_sink.pipe. Each day at 00:30 UTC it exported the v1 health score to the health_score_sink Kafka topic. The v2 export, health_score_v2_sink, already carries the v2 fields this pipe exported.
• The pipe exists only in prod. It uses the prod-only lfx-oracle-kafka-streaming connection, so staging never had it.
• lf-dbt reads this export through its health_score_sink source. lf-dbt PR 2 moves the v2 columns to the v2 sink, and lf-dbt PR 4 removes the source. This PR waits for both.
• Removes only the pipe. The datasource it reads, health_score_copy_ds, stays until crowd.dev PR 3. The Snowflake tables and the connector topic are human steps below.

Deploy notes

Prerequisites

• lf-dbt PR 4 is merged and has run in prod. Done: linuxfoundation/lf-dbt#3036 and linuxfoundation/lf-dbt#3037 merged on 2026-10-19, and lf-dbt main no longer reads health_score_sink.
• Consumer sign-off is no longer needed (decided 2026-10-13).

Steps

Prod only. Staging never had this pipe.

tb pipe rm health_score_sink --yes

1. Confirm the pipe is gone on prod with tb pipe ls.
2. The next day, confirm no export ran at 00:30 UTC.
3. Then the human steps: remove health_score_sink from the Snowflake sink connector (crowd-kube connector runbook), confirm the connector tasks are RUNNING, and delete the OCI stream health_score_sink.

Cross-repo order

1. lf-dbt refactor: search endpoint sorting improvements #3034, refactor: async syncronize github nango integration with nango cloud api #3035, feat: implement API cache for the project groups page #3036, fix: security best practices rename tables to conform to kafka topic limits #3037 (merged)
2. crowd.dev PR 2 (this PR)
3. crowd-kube connector change (human)
4. Snowflake DROPs (human)

On hold until org-dashboard retires (expected early November 2026), because it still reads the v1 score through /api/project?healthScore=true: insights #2375 and #2376, crowd.dev #4922 and #4923 (closed for now, branches kept).

What breaks if out of order

• Before lf-dbt PR 4 has run in prod: the health_score_sink source freshness check (error_after: 1 day) fails the dbt Cloud job, and the silver health leg goes stale.
• Before lf-dbt PR 2: lf-dbt's v2 columns still come from this export. Removing it sets them to NULL. The LFX One drawer shows Unscored for every project, and the project recent-progress card shows 0.
• Deleting the OCI stream before the connector change: one Snowflake sink connector carries every topic. A missing topic fails its task and halts every Snowflake sink, including the v2 sink.

Related PRs

• crowd.dev PR 1: add health score v2 to project lists and re-source dependency metrics (live)
• crowd.dev PR 3: remove v1 health score computation and columns from Tinybird (gated)
• crowd.dev PR 4: remove the v1 repository health score pipes (gated)
• insights PR 1: show health score v2 in the open source index
• insights PR 2: remove health score v1 fields and routes (gated)
• lf-dbt PR 1: expose the v2 sink leg marker on gold project health metrics
• lf-dbt PR 2: source health score v2 columns from the v2 sink
• lf-dbt PR 3: drop v1 health score columns from gold, platinum and silver dim (gated)
• lf-dbt PR 4: remove the v1 health score sink source and silver leg (gated)
• lfx-self-serve PR (This is a test issue #3465): remove the v1 health band shim
• crowd-kube connector change (gated, human step): no PR exists yet. In the crowd-kube connector runbook kept with IN-1222, a human removes the topic from the Snowflake sink connector and opens that PR. crowd-kube Add links to dashboard sentiment widget #205 changes docs only.

JIRA

IN-1222: Deprecate v1 health score sinks and tables
https://linuxfoundation.atlassian.net/browse/IN-1222

[Copilot]

🔵 Needs a closer look

Merge depends on external production sign-off and deprecation prerequisites that cannot be verified from this diff.

0 open findings

What changed in this PR

Removes the deprecated v1 health-score Kafka export after downstream migration to v2.

Changes:

• Deletes the production-only health_score_sink Tinybird pipe.
• Retains its datasource and v2 sink.

File	Description
services/​libs/​tinybird/​pipes/​health_score_sink.pipe	Removes the scheduled v1 Kafka sink.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[cursor]

PR Summary

Medium Risk
Stops a prod Kafka export relied on by the v1 health score pipeline; safe only after lf-dbt and other consumers have migrated to the v2 sink.

Overview
Removes the prod-only Tinybird pipe health_score_sink, which previously ran daily at 00:30 UTC and exported v1 health score fields (plus some v2 columns from project_insights_copy_ds) to the health_score_sink Kafka topic via lfx-oracle-kafka-streaming.

After deploy, tb pipe rm health_score_sink stops that daily CSV export; downstream consumers are expected to use health_score_v2_sink instead (lf-dbt no longer sources this topic). Related datasources like health_score_copy_ds and Snowflake/OCI cleanup are out of scope for this diff.

^{Reviewed by Cursor Bugbot for commit 12e840a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

🔵 Needs a closer look

Production removal requires human confirmation that the prerequisite lf-dbt changes have run successfully in production.

No open issues 🎉

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Human confirmation is needed that production lf-dbt prerequisites are satisfied and the cross-repository deployment sequence is safe.

No open issues 🎉

_{Review effort: Balanced}