Some DLTpipelines suddely seem to take different runtime 16.1 instead of 15.4 since last night (CET)

[cookiebaker]

Hello, 

Suddenly since last night on some of our DLT pipelines we're getting failures saying that our hive_metastore control table cannot be found. All of our DLT's are set up the same (serverless), and one Shared Compute on runtime version 15.4. 

For the failing pipelines, when looking at the Update Details - > Logs -> Configuration tab, that the failed pipelines take runtime "dlt:16.1.5-delta-pipelines-photon-dlt-release-dp-2025.15-rc0-commit-be1adbe-image-0e56918". 

The pipelines that succeeded all just took the "dlt:15.4.12-delta-pipelines-photon-dlt-release-dp-2025.15-rc0-commit-be1adbe-image-5055805". (starting with 15.4). 

Did something change on the Databricks end? For us nothing changed in the settings and seems like a sudden disruption of DLT pipelines that were previously just running succesfully.

Thank you in advance.

Best, Nigel

[BigRoux]

This kind of issue might arise due to the automatic upgrade of the runtime version within the serverless environment, as some updates introduce changes potentially affecting dependencies or configurations. Databricks runtime upgrades typically include bug fixes, enhancements, and new features, which might lead to compatibility issues with existing pipelines if the code or configuration depends on the previous runtime behavior.

Serverless is a fully managed service which includes runtime managment.

[cookiebaker]

Thank you for your message. 

Where can I find the release notes of this latest change that caused the disruption?

We were forced to move our DLT's back to the normal computes and moved them away from serverless.

[voo-rodrigo]

We are also experiencing the same issue. Since around 6pm ET last night (04-22) our Serverless DLT Pipelines cannot read the hive_metastore. It doesn't seem to be related to Databricks Runtime Version 16.1, however, because previous runs using the same version were successful. Also, we are able to run these DLTs using managed Clusters instead of Serverless, even if the managed Clusters are also at version 16.1. Also, when running regular notebooks or job notebooks using Serverless, we are able to query the hive_metastore as well. The issue seems to only affect DLT Serverless.

We wouldn't want to switch all our pipelines to stop using Serverless, but that is the only setup that currently works for us. Could anything have changed last night specifically at DLT Serverless configuration?

[JeanSeb]

any update on this one? we're also experiencing the same kind of issue since yesterday evening, with serverless DLT referring to hive metastore tables

[BigRoux]

Serverless release notes can be found here:  https://docs.databricks.com/aws/en/release-notes/serverless/

[voo-rodrigo · Accepted solution]

Hello. FYI last week I opened a ticket with Databricks Support. They identified a flag that had been deployed last week as the culprit, and rollback the changes. I tested it this morning and I am now able to read the hive_metastore from Serverless DLTs. Thank you.

[cookiebaker]

@voo-rodrigo Hello, thanks for updating the progress on your end! I've tested as well and confirmed that the DLT can read the hive_metastore via Serverless again.