How to Refresh Materialized Lake Views in Microsoft Fabric

Two ways to refresh a materialized lake view in Microsoft Fabric, from code in your ELT process or on a schedule, and when to use which.

In this article 4 sections
  1. Why a materialized lake view needs a refresh
  2. Option 1: refresh from code
  3. Option 2: refresh on a schedule
  4. Which one to use

A materialized lake view (MLV) in Microsoft Fabric stores its data physically on the lakehouse. When new data lands in its source tables, the view does not pick that up by itself, so you have to refresh it. Fabric gives you two ways to do that. You can refresh from code, as part of the process that loads your data, or you can put the view on a schedule. Which one you want depends on whether you know when new data comes in.

For the examples I use the view I built when I showed how to create materialized lake views: vw.CountByBorough, which counts the rows per borough in the taxi zone lookup table.

How to refresh materialized lake views in Microsoft Fabric (5 min) Watch on YouTube

Why a materialized lake view needs a refresh

The definition of the view sits in a Spark SQL cell in the notebook where I created it:

SQL
%%sql
create materialized lake view vw.CountByBorough as
select Borough, count(*) as RowCount
from raw.TaxiZoneLookup
group by Borough

It reads raw.TaxiZoneLookup and counts the rows for each borough, and the result shows up as a table on the lakehouse.

What is very important to remember is that the view contains physical data on the lakehouse. That makes it quite different from a view in SQL Server. The result of the query is stored, and it only changes when you refresh it, which is something you have to do yourself.

Option 1: refresh from code

The first way is to run a Spark SQL command: refresh materialized lake view, followed by the schema and the name of the view.

SQL
%%sql
refresh materialized lake view vw.CountByBorough

What you get back is a metadata overview of the refresh, with one metric per row. Mine says Fabric did a full refresh on 31 July 2025 and dropped no rows.

Notebook cell running refresh materialized lake view vw.CountByBorough, with a result table of metrics such as Namespace, MLVName, RefreshPolicy FullRefresh, RefreshDate and TotalRowsDropped 0
The refresh command returns a metadata overview of the refresh

Why I like this option

To me this is one of the better solutions, because you can put it into a metadata-driven processing framework. You load data into your tables through metadata, and the same ETL process then refreshes the materialized lake views on top of those tables at the point where new data has actually come in.

In practice that means a notebook that goes through a list of the available materialized lake views (or some other configuration metadata) and refreshes them one by one. If you are new to metadata-driven processing, I explained the idea in Metadata Driven Ingestion into Fabric with Pipelines.

Option 2: refresh on a schedule

Sometimes you do not control when data comes in. It just trickles in whenever, outside of your process, and there is no load step to hang a refresh command on. In that case you want the view refreshed on a set interval so it is kept up to date anyway.

Open the materialized lake view settings

Go to your lakehouse and find the view in the explorer. Mine lives in the vw schema. Then click Manage materialized lake views (preview) in the ribbon.

Lakehouse explorer with vw.countbyborough selected and the Manage materialized lake views (preview) button in the ribbon
Manage materialized lake views sits in the lakehouse ribbon

This opens the lineage of your materialized lake views, which shows you where the data comes from. In my case vw.countbyborough reads from one single table, raw.taxizonelookup. The dropdown in the top left is brilliant, by the way, because the only option in it is View lineage. A bit odd. I would not have guessed that.

Manage materialized lake views page showing lineage from raw.taxizonelookup to vw.countbyborough, with Run and Schedule buttons in the toolbar
The lineage view, with the Schedule button in the toolbar

Set the schedule

Click Schedule in the toolbar and set Schedule refresh to On. Under Repeat you choose how often it runs: by the minute, hourly, daily, weekly or monthly. I picked daily at 15:00. Below that you set a start date and time (today, in my case), an end date a couple of years out, and the time zone.

Schedule pane with Schedule refresh set to On, Repeat set to Daily and Time set to 15
A daily refresh at 15:00

When you click Apply, the view gets its refresh schedule. From then on the schedule refreshes it whether there is new data in the source tables or not.

Which one to use

Refresh from codeRefresh on a schedule
Use it whenYour own ETL process loads the source tablesData comes in outside of your control
Howrefresh materialized lake view in Spark SQL, for example in a notebookSchedule under Manage materialized lake views
When it refreshesRight after new data has landedOn a fixed interval, with or without new data

If you load the source tables yourself, refresh from code and let your metadata decide which views to refresh. If you do not know when data arrives, put the view on a schedule.

Bas Land · Microsoft Data Platform MVP

Bas has spent more than 12 years in the Microsoft data stack. He builds data platforms on Microsoft Fabric, teaches as a Microsoft Certified Trainer and speaks at conferences such as Power BI Gebruikersdagen.

↑↓to move ↵to open escto close