Skip to content

Search the site

Microsoft retires SQL Server 2019 Big Data Clusters early

"This is a big shift, and it's kind of concerning..."

Microsoft has unexpectedly announced the retirement of SQL Server 2019 Big Data Clusters -- an offering that lets users deploy scalable SQL Server, Apache Spark, or Hadoop Distributed File System containers on Kubernetes on-premises or in the cloud. The "quite significant" decision to kill off this feature of an enterprise product by 2025 has ruffled the feathers of some users given SQL Server 2019 doesn't leave Extended support until 2030.

The decision came in a February 25 update on one of Microsoft's sprawling suite of product blogs. Redmond also announced that Cloudera (CDP) and Hortonworks (HDP) external data sources will be retired for all in-market versions of SQL Server and not be included in SQL Server 2022. The move  suggests a continued drive to get customers running analytics workloads onto Azure/more up-to-date tools, but sunk costs mean those who have taken the time and effort to set up Big Data Clusters may not be best pleased with the early death of support.

In a post attributed to the SQL Server team"Microsoft said the PolyBase scale-out groups in Microsoft SQL Server feature will also be killed off in SQL Server 2022. (PolyBase is a data virtualisation feature for SQL Server that lets users query data from a range of sources without installing client connection software. SQL Server 2019, 2017, and 2016 will support the functionality to the end of support -- 2030, 2027, and 2026 respectively.)

Goodnight, SQL Server 2019 Big Data Clusters

Wes Miller, a former Microsoft product manager and now a research analyst for independent IT advisory service Directions on Microsoft, said: " I discovered this change because the new product terms on March 1 noted without any further elaboration, 'Removed SQL Server Big Data Nodes offer for service end-of-life.'

He added on Twitter: "Microsoft is (quite unusually, in my experience) removing a (highly touted) feature as soon as Mainstream support ends, but half of the lifecycle before the product leaves support entirely.

"This is a big shift, and it's kind of concerning, esp. since the feature required SA [Software Assurance: A software maintenance programme from Microsoft]... This means that customers should treat any subscriptified [sic] enterprise product (or SA-mandated features within such a product) with care, as the rules (the guidelines?) are apparently becoming less well-defined, and squishier. Data Nodes... was a feature of an enterprise product, which required SA," he noted, adding in a thread: "Also, it's weird that the blog post called it an "add-on".

"Big Data Clusters was not an add-on. That's a reserved term in licensing.

"There was a way to add additional core licenses to the cluster if needed, but it wasn't the only way to license it, or the default way, You licensed it by way of Enterprise cores with SA... It was needlessly complex, and we wasted a tonne of time (for my colleague) on trying to distill down how it worked into some consumable form."

Microsoft said: "All existing users of SQL Server 2019 with Software Assurance will be fully supported on the platform for the next three years, through February 2025... In the latest version of SQL Server, we are engineering the best mix of on-premises and in-cloud relational workloads and connectivity to Azure Synapse Analytics for advanced analytics in a flexible, scalable, and integrated environment."