skills.rest
airbytehq
airbytehqOFFICIAL
Explore all skills by this author →
Website𝕏@AirbyteHQGitHub

source-mssql-e2e-cdc-tests

Reproduce CDC bugs against source-mssql using SQL fixtures and declarative test assertions.

22.0k|5.3k|Updated Jul 27, 2020
Search Tags:#SQL Server#bug reproduction#E2E Testing#cdc#debezium#airbyte#source-mssql
One-click install
npx skills add https://github.com/airbytehq/airbyte --skill source-mssql-e2e-cdc-tests
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: source-mssql-e2e-cdc-tests
Source: https://github.com/airbytehq/airbyte/tree/main/airbyte-integrations/connectors/source-mssql/.agents/skills/source-mssql-e2e-cdc-tests
Command: npx skills add https://github.com/airbytehq/airbyte --skill source-mssql-e2e-cdc-tests

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reproducing customer-reported CDC bugs in the Airbyte source-mssql connector locally is slow and error-prone: you must enable CDC on SQL Server, craft per-bug database state, run the connector, and verify failure signatures by hand. This Skill packages that entire loop into runnable case scripts with declarative pass/fail assertions.

Core Features & Use Cases

  • Per-bug case scripts: Ready-made reproductions for airbytehq/oncall#11451 (LSN-range regression), #12094 (schema-history bloat), and #12162 (whitespace in stream names), each gating its exit code on --expect-* assertions.
  • CDC fixture library: Idempotent SQL fixtures, CDC-aware config templates, and configured catalogs that work across source-mssql 4.3.x and 4.4.x.
  • Multi-phase repro support: Capture Airbyte STATE messages between reads, mutate the server (e.g., advance the CDC min LSN), and replay with stale state to test offset-validation behavior.
  • Use Case: After building a fix with :dockerBuildx, run VERSION=dev ./cases/11451.sh to confirm the fix still rejects genuinely expired saved offsets while resolving the reported bug.

Quick Start

Start the SQL Server backend with the generic skill's start-backend.sh, then run the case script for the bug you want to reproduce, such as cases/12162.sh.

Frequently Asked Questions about source-mssql-e2e-cdc-tests

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reproduce a source-mssql CDC bug locally?▼

Start the SQL Server backend with the generic skill's start-backend.sh, then run the case script for your bug, such as cases/12162.sh. Each case applies its SQL fixtures, renders the CDC config, runs the connector, and fails with a non-zero exit code if its --expect-* assertions are not met.

How do I verify a source-mssql CDC fix before releasing it?▼

Build the connector locally with ./gradlew :airbyte-integrations:connectors:source-mssql:dockerBuildx, then re-run the relevant case with VERSION=dev, for example VERSION=dev ./cases/11451.sh. The case's assertions confirm the fix resolves the bug without breaking offset validation.

Does this skill require Airbyte Cloud or GSM credentials?▼

No. It runs entirely locally against a Dockerized SQL Server backend container. You only need Docker, uv, jq, and a clone of the airbytehq/airbyte repository.

How do I test a multi-phase CDC bug involving stale state?▼

Model your case on cases/11451.sh: run a baseline read, extract the STATE messages with extract-state.py, mutate the server with apply-sql.sh, then replay with run.sh --state=PATH and --skip-fixtures. Use --step-name=<bug>/<phase> to keep each phase's artifacts separate.

Why does source-mssql fail with a Debezium message.key.columns error?▼

This happens when a CDC-enabled stream's identifier contains whitespace, because buildMessageKeyColumns() joins schema.table:pkcol strings without filtering or escaping them. Case 12162 reproduces this; the fix pre-filters such streams so Debezium falls back to the table's native primary key.

Can I run the CDC cases against older connector versions?▼

Yes. Cases default to VERSION=4.4.2 but accept overrides, for example VERSION=4.3.4 ./cases/12162.sh. Case 11451 additionally splits BASELINE_VERSION and TARGET_VERSION so you can capture state on one version and replay on another.

Related Skills

Explore More Software Engineering Skills→
SnapLogicSnapLogic

setup-database

Manage database containers for SnapLogic pipeline testing across multiple database types.

Official
Advanced
curdxcurdx
4

verification-before-completion

Reproduce failures before fixes and confirm resolution with documented evidence.

Community
Advanced
anton-abyzovanton-abyzov
156

confluent-kafka-connect

Configure and manage Kafka Connect source and sink connectors.

Community
Advanced
tensormuxtensormux
54

Optimize Triton Block Parameters

Autotune Triton GEMM block sizes and launch parameters for GPU throughput.

Official
Advanced