Technology1 min read

Java4Shell vulnerability (CVE-2026-114938) assessment

By · Published by Everything Blog

In short

Databricks has acted to secure its environments against an industry-wide OpenJDK flaw by implementing a two-phase defense strategy. To immediately disable the vulnerable code across all supported runtime versions, the company took the unusual step of proactively restarting affected clusters. Full security will be achieved in the second phase through the deployment of patched runtime versions to permanently replace the defective code. Serverless platform users will get these patches automatically, while classic cluster users must restart systems and custom image operators will receive direct outreach.

Key points

  • The initial mitigation to disable vulnerable code is not anticipated to create a meaningf…: The initial mitigation to disable vulnerable code is not anticipated to create a meaningful performance impact for customers.
  • Customers operating classic clusters will need to restart them manually once permanent ru…: Customers operating classic clusters will need to restart them manually once permanent runtime patches are released.
  • Serverless environments will receive complete patches automatically, while Databricks is…: Serverless environments will receive complete patches automatically, while Databricks is directly contacting users running custom images.

Sources checked by our analysis pipeline: 1

by Fermín J. Serna

As of November 19 at 10 AM Pacific Time:

Prior to Java4Shell’s public disclosure (CVE-2026-114938), Databricks was already addressing the industry-wide OpenJDK vulnerability.

Databricks is protecting customers in two phases: quick mitigation to disable use of the vulnerable code, and then patching to replace the vulnerable code.

- Mitigation: We have applied the mitigation to our platform, disabling use of vulnerable code for all supported versions of Databricks Runtime. We do not expect the mitigation to have a meaningful performance impact on customers. In order to apply the mitigation as quickly as possible, we have taken the rare step of proactively restarting affected clusters.

- Patching: While the mitigation blocks the attack, in an abundance of caution, we will be providing patched versions of supported Databricks Runtimes. Customers on our serverless platform will automatically get them. Customers with classic clusters will need to restart their clusters.

We are also reaching out to customers running custom images.

We have no evidence to indicate any unauthorized access to Databricks products, services, or customers' data.

Technology