Synapse Link for Dataverse: Upgrading Your Synapse Spark Pool Version
How to perform an in-place Synapse Spark pool upgrade for Synapse Link for Dataverse. Originally written for the 3.3 to 3.4 move — updated for Spark 3.5 and Delta Lake 3.0, since Spark 3.4 is now retired.
Updated September 2026. This post was written for the Spark 3.3 → 3.4 upgrade. Spark 3.4 is now itself retired.
- The current requirement for Synapse Link Delta Lake profiles is Apache Spark 3.5 with Delta Lake 3.0.
- Retired runtimes can't be used for new Spark pools, and existing workflows on them won't execute.
- The upgrade procedure below is unchanged — create a new pool on the current version, then upgrade the profile in place. Substitute 3.5 wherever this post says 3.4.
Before you invest in this upgrade, note that the Delta Lake (Parquet) export option in Synapse Link is deprecated — closed to new customers from 15 October 2026 and removed by December 2027. Upgrading keeps you running in the meantime, but it isn't a destination. See Azure Synapse Link Delta Lake Is Being Retired for the migration decision.
Microsoft Dynamics suite is a widely used product for implementing ERP and CRM solutions. Synapse Link for Dataverse is a functionality to extract data from Microsoft dynamics systems for analytics and integration purposes.

In this article, We will look into how to upgrade Synapse spark pool version to 3.4 in Synapse link for dataverse.
#What is Synapse Link for Dataverse?
Synapse Link is a feature within Dataverse designed to efficiently manage and export data from microsoft Dynamics systems to external systems such as Azure Data Lake (CSV or delta) or Microsoft Fabric.

#Export data using synapse link in Delta format
When setting up a connection to extract data from Dynamics 365 to Azure Data Lake in Delta format, it's essential to configure a Synapse Spark pool during the Synapse Link setup. This step ensures that the data is processed efficiently and stored in the optimal Delta format within your Azure Data Lake.

Synapse Spark pool is used to convert data from CSV to Delta format in Data lake.
With support for Synapse Spark Pool version 3.3 coming to an end on 31st March 2025, it's crucial to upgrade your existing Synapse Link to Spark Pool version 3.4. This upgrade will ensure continued support, enhanced performance, and access to the latest features, keeping your data integration processes running smoothly. By making the switch ahead of time, you'll avoid potential disruptions and ensure your system remains optimized and future-ready
September 2026 update: Spark 3.4 has since been retired as well. The current version to target is Spark 3.5 with Delta Lake 3.0. The same reasoning and the same upgrade steps apply — only the version number has moved on.

#How to upgrade synapse spark version to 3.4?
- First step is to create a new synapse spark pool with version 3.4 in synapse workspace

- Second step is to upgrade the existing synapse link in dataverse environment.

Pro tips:
- Here is a link to microsoft documentation.
- The Spark pool upgrade occurs only when a new Delta Lake conversion Spark job is triggered — so make sure at least one data change happens after you select Update.
- You must create a new Synapse Spark pool with the current Spark version, using the same or higher nodes hardware configuration within the same Synapse workspace.
- Create the new pool independently of the existing one. Don't delete your current Spark pool, and don't give the new pool the same name.
- After you select Update, the upgrade can take up to 48 hours to take effect because of caching. The old pool keeps handling Delta Lake conversion until then — don't delete it until you've confirmed the new pool is being used.
#Related reading
- Azure Synapse Link Delta Lake Is Being Retired — why this upgrade path has an expiry date, and what to move to.
- What Link to Microsoft Fabric Actually Costs — if you're considering Link to Fabric as the replacement, this is the cost model to check first.
Get future articles
Follow for practical Microsoft Fabric, Azure, Spark, and data engineering writeups.