Pipeline Migration

Telemetry — MANDATORY. Every api.fabric.microsoft.com call must carry x-ms-fabric-skill: pipeline-migration (az rest: --headers "x-ms-fabric-skill=pipeline-migration"), including every LRO poll, fabric_lro and retry. Snippets omit it — add it anyway.

CRITICAL NOTES

1. To find workspace details (including its ID) from a workspace name: list all workspaces, then use JMESPath filtering
2. To find item details (including its ID) from workspace ID, item type, and item name: list all items of that type in that workspace, then use JMESPath filtering
3. Fabric does NOT have a "Dataset" item type — all dataset properties are inlined into activity typeProperties
4. Linked Services map to Fabric Connections — in pipeline activity JSON, referenceName on the activity's linkedService block uses the Fabric connection display name (not the connection GUID); the connection GUID is used only on Fabric REST API calls
5. Notebook activities change from SynapseNotebook to TridentNotebook and reference notebooks by GUID, not by name
6. Synapse global parameters become a Variable Library item in Fabric, referenced as @pipeline().libraryVariables.<name>. Variable Library Number types are not consumable in pipelines — Synapse Float/Double are mapped to String for runtime compatibility
7. The Validation activity type does not exist in Fabric — it must be rewritten as GetMetadata + IfCondition
8. Triggers are intentionally excluded from this skill — recreate schedules manually in Fabric after migration
9. SSIS package execution, SHIR-exclusive connectors, and Databricks activities are parked — see pipeline-gotchas.md

Synapse Pipelines → Microsoft Fabric Data Factory Migration

Prerequisite Knowledge

These companion documents provide general Fabric REST patterns. Do NOT read them upfront — reference only when a specific phase requires a pattern not already covered in this skill's resource files:

• COMMON-CORE.md — General Fabric REST API patterns, authentication & token audiences, item discovery via JMESPath
• COMMON-CLI.md — az rest / az login CLI patterns, authentication recipes, pipeline run/schedule operations
• ITEM-DEFINITIONS-CORE.md — DataPipeline and VariableLibrary item definition structures (pipeline-content.json, variables.json)
• SPARK-AUTHORING-CORE.md — Fabric notebook item creation (needed when notebook items don't exist yet in Fabric)

For the notebook side of the migration (the notebooks that pipeline activities call), use the companion synapse-migration skill to migrate the notebook content itself.

Table of Contents

Context Loading Guide

IMPORTANT — Load only what you need. Do NOT read all resource files upfront.

API-Driven Migration Workflow

Authentication

Use az account get-access-token --resource <audience> --query accessToken -o tsv to acquire tokens.

Synapse Data-Plane API Reference

Fabric API Reference

Assessment Mode (Optional but Recommended)

Before creating any items in Fabric, run the pipeline assessment to understand scope, complexity, and blockers. The assessment is read-only — it queries Synapse APIs only and produces a markdown report.

Copilot workflow — no Python script file needed:

1. Ask the user: "What is the name of your Synapse workspace?"
2. Auto-discover subscription ID and resource group via az account show and az synapse workspace show
3. Run the assessment code inline from the terminal
4. Print the full report output directly in the chat

To also save the report to disk, pass output_path=f"pipeline-assessment-{SYNAPSE_WS}.md" to generate_assessment_report(). The PipelineAssessment objects it produces feed directly into the migration scripts in pipeline-orchestrator.md.

Migration Mode (Inline — No Script File Required)

Copilot performs the migration directly from the terminal. No Python files to save or run manually.

Ask the user for:

1. Synapse workspace name (reuse from assessment if already run)
2. Fabric workspace name
3. Which pipelines to migrate — specific names, or * for all
4. Optional name suffix to append to each pipeline in Fabric (e.g. _migrated) — leave blank to keep the original name

Everything else is auto-discovered:

• Subscription ID and resource group via az account show + az synapse workspace show
• Fabric workspace ID via GET /v1/workspaces filtered by display name
• Notebook GUIDs in Fabric via GET /v1/workspaces/{wsId}/notebooks (required for SynapseNotebook → TridentNotebook)
• Connection names in Fabric via GET /v1/connections (when datasets reference linked services)

What is fully automated inline:

• ✅ SynapseNotebook → TridentNotebook — type rename, GUID lookup, remove sparkPool/sessionConfiguration, fix timeout to 12h max
• ✅ All compatible activity types — pass through with minor property adjustments
• ✅ Dataset inlining into activity typeProperties
• ✅ Global parameter expressions rewritten to @pipeline().libraryVariables.<name>
• ✅ Pipeline JSON assembly and deployment to Fabric via REST

Copilot checks before starting — will pause and report if:

• A notebook referenced by a SynapseNotebook activity does not exist in the Fabric workspace yet
• A Fabric Connection is missing for a linked service referenced by dataset activities

Load pipeline-orchestrator.md for the complete inline runner.

Migration Phases (Execute in Order)

Phase 0 must precede Phase 4: Fabric Notebook GUIDs are needed before TridentNotebook activities can be written. Phase 2 must precede Phase 3: Connection names must exist before they can be referenced in inlined datasets.

Activity Type Quick Reference

Full mapping table, before/after examples, and parking decisions are in activity-mapping.md.

Must / Prefer / Avoid

MUST DO

• Migrate notebooks before pipelines — Fabric TridentNotebook activities require notebook GUIDs, not names. Use the synapse-migration skill first
• Create Fabric Connections before building pipeline JSON — linked service names in Synapse become connection references in Fabric; you need the connection names before inlining datasets
• Inline all dataset definitions — Fabric Data Factory has no Dataset item type; all inputs/outputs dataset properties must be embedded in each activity
• Replace @pipeline().globalParameters.<name> with @pipeline().libraryVariables.<name> after creating the Variable Library
• Replace Validation activities with a GetMetadata + IfCondition pair — Validation does not exist as an activity type in Fabric
• Remove sparkPool and sessionConfiguration from migrated notebook activities — pool selection and session config belong in the Fabric Environment attached to the notebook

PREFER

• Variable Library with Value Sets for dev/test/prod environments — use @pipeline().libraryVariables.<name> with environment-specific Value Sets instead of pipeline-level parameters for environment promotion
• Parameterized notebookParameters over hardcoded values in TridentNotebook activities — mirrors Synapse parameterized notebook pattern
• Test notebook activities individually before running the full migrated pipeline — notebook GUIDs are the most common source of failure
• OneLake Lakehouse sources/sinks for Copy activities where applicable — eliminates the need for external connections for data already in OneLake

AVOID

• Do not reference notebooks by name in TridentNotebook activities — use the GUID from the Fabric workspace notebook list
• Do not carry over Synapse triggers — they are not migrated by this skill; recreate schedules in Fabric after validating the pipeline
• Do not attempt to migrate SSIS, Databricks, or AzureBatch activities without reading pipeline-gotchas.md first — these require manual intervention
• Do not hardcode workspace/item GUIDs in pipeline JSON — use Variable Library entries or pipeline parameters so environments can be promoted without editing pipeline JSON
• Do not use @pipeline().globalParameters syntax after migration — this expression path does not exist in Fabric; all migrated global parameters must be accessed via @pipeline().libraryVariables

Migration Gotchas — Quick Reference

Full troubleshooting guide is in pipeline-gotchas.md.

Post-Migration: What's Next

After pipeline migration, hand off to these companion skills and tools:

Examples

SynapseNotebook → TridentNotebook activity (before/after)

{
  "name": "Run_Notebook",
  "type": "SynapseNotebook",
  "typeProperties": {
    "notebook": {"referenceName": "MyNotebook", "type": "NotebookReference"},
    "sparkPool": {"referenceName": "BigPool", "type": "BigDataPoolReference"}
  }
}

After migration (GUID from GET /v1/workspaces/{wsId}/notebooks):

{
  "name": "Run_Notebook",
  "type": "TridentNotebook",
  "typeProperties": {
    "notebookId": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
    "workspaceId": "yyyyyyyy-yyyy-yyyy-yyyy-yyyyyyyyyyyy"
  }
}

Global parameter expression rewrite

Before (Synapse):  @pipeline().globalParameters.batchDate
After (Fabric):    @pipeline().libraryVariables.batchDate

Dataset inlining: Copy activity — dataset reference → connection display name

Synapse dataset InputBlobDataset (will be inlined and removed):

{
  "type": "AzureBlob",
  "linkedServiceName": {"referenceName": "AzureBlobLinkedService"},
  "typeProperties": {"folderPath": "input/", "fileName": "data.csv"}
}

Synapse Copy activity (before — references dataset by name):

{
  "name": "CopyData", "type": "Copy",
  "inputs": [{"referenceName": "InputBlobDataset", "type": "DatasetReference"}],
  "typeProperties": {"source": {"type": "BlobSource"}, "sink": {"type": "BlobSink"}}
}

After migration (connection display name for AzureBlobLinkedService is My ADLS Connection):

{
  "name": "CopyData", "type": "Copy",
  "typeProperties": {
    "source": {"type": "BlobSource", "storeSettings": {"type": "AzureBlobStorageReadSettings"}},
    "sink":   {"type": "BlobSink",   "storeSettings": {"type": "AzureBlobStorageWriteSettings"}}
  },
  "linkedService": {"referenceName": "My ADLS Connection", "type": "LinkedServiceReference"}
}

See activity-mapping.md and notebook-activity-migration.md for full before/after examples.