Fabric Materialized Lake Views Performance remediate

Diagnose and resolve performance issues with materialized lake views (MLVs) in Microsoft Fabric lakehouses. This skill covers refresh optimization, Spark job diagnostics, data quality constraint tuning, and lineage execution remediate.

When to Use This Skill

• •MLV refresh runs are taking longer than expected
• •Incremental refresh is falling back to full refresh unexpectedly
• •MLV lineage execution shows Failed or Skipped nodes
• •Spark jobs for MLV refresh are failing with errors
• •"Delta table not found" errors during MLV creation or refresh
• •Data quality constraints causing unexpected pipeline failures
• •Need to enable or verify optimal refresh configuration
• •Custom Spark environment tuning for MLV workloads
• •Monitoring and interpreting MLV run history

Prerequisites

• •Microsoft Fabric workspace with Lakehouse items
• •Schema-enabled lakehouse (recommended for MLV support)
• •Fabric notebook for executing Spark SQL commands
• •Workspace Admin or Contributor role for scheduling and monitoring
• •Access to Monitor Hub for viewing MLV run details

Quick Diagnostics Checklist

Run through these checks in order when remediate MLV performance:

Step-by-Step Workflows

Workflow 1: Diagnose Slow Refresh

1. •Determine current refresh strategy - Navigate to Manage materialized lake views in the lakehouse ribbon. Check if Optimal refresh toggle is ON.
2. •Verify change data feed - Run the diagnostic script to check CDF status on all source tables. See scripts/check-cdf-status.sql.
3. •Check for unsupported expressions - Review the MLV definition for constructs that force full refresh. See Supported Expressions.
4. •Inspect source data patterns - If source tables have UPDATE or DELETE operations, Fabric always performs full refresh regardless of CDF status.
5. •Review partition strategy - Consider adding PARTITIONED BY to large MLVs to improve refresh parallelism.
6. •Attach custom environment - Configure a custom Spark environment with optimized compute for heavy workloads. See references/custom-environment-guide.md.

Workflow 2: Resolve Failed Refresh Runs

1. •Open lineage view - Navigate to Manage materialized lake views, select the failed run from the dropdown (last 25 runs available).
2. •Identify failed node - Click the failed node in the lineage graph. Review the error message in the right-side panel.
3. •Access Spark logs - Click the Detailed Logs link to navigate to Monitor Hub. Review Apache Spark error logs.
4. •Common failure patterns - See references/common-failure-patterns.md for resolution steps.
5. •Retry or recreate - For transient Spark failures, retry the run. For persistent errors, drop and recreate the MLV.

Workflow 3: Enable Optimal Refresh

1. •Enable CDF on all source tables:

ALTER TABLE bronze.customers SET TBLPROPERTIES (delta.enableChangeDataFeed = true);
ALTER TABLE bronze.orders SET TBLPROPERTIES (delta.enableChangeDataFeed = true);

1. •Verify CDF is enabled:

DESCRIBE DETAIL bronze.customers;
-- Check properties column for delta.enableChangeDataFeed=true

1. •Create MLV with CDF property:

CREATE OR REPLACE MATERIALIZED LAKE VIEW silver.customer_orders
TBLPROPERTIES (delta.enableChangeDataFeed=true)
AS
SELECT 
    c.customerID,
    c.customerName,
    c.region,
    o.orderDate,
    o.orderAmount
FROM bronze.customers c INNER JOIN bronze.orders o
ON c.customerID = o.customerID;

1. •Enable optimal refresh toggle - Navigate to Manage materialized lake views and verify the Optimal refresh toggle is ON (enabled by default).

Supported Expressions for Incremental Refresh

MLVs using only these constructs qualify for incremental refresh:

Unsupported constructs that force full refresh:

• •LEFT JOIN, RIGHT JOIN, FULL OUTER JOIN
• •Window functions (ROW_NUMBER, RANK, LAG, LEAD, etc.)
• •Non-deterministic functions (current_timestamp(), rand(), etc.)
• •Subqueries in SELECT or WHERE
• •GROUP BY with HAVING
• •DISTINCT
• •User-defined functions (UDFs)

Key Spark SQL Reference

List All MLVs

SHOW MATERIALIZED LAKE VIEWS IN silver;

View MLV Definition

SHOW CREATE MATERIALIZED LAKE VIEW silver.customer_orders;

Force Full Refresh

REFRESH MATERIALIZED LAKE VIEW silver.customer_orders FULL;

Drop and Recreate

DROP MATERIALIZED LAKE VIEW silver.customer_orders;

CREATE OR REPLACE MATERIALIZED LAKE VIEW silver.customer_orders
TBLPROPERTIES (delta.enableChangeDataFeed=true)
AS
SELECT ...;

Known Issues and Limitations

Run States Reference

remediate

References

• •Common Failure Patterns - Detailed resolution for frequent MLV errors
• •Custom Environment Guide - Spark environment tuning for MLV workloads
• •Monitoring Deep Dive - Advanced Monitor Hub usage for MLV diagnostics
• •CDF Status Check Script - T-SQL/Spark SQL script to audit CDF property
• •MLV Health Check Script - Comprehensive MLV configuration audit
• •MLV Diagnostic Template - Notebook template for systematic MLV remediate