/
3.1 Release Notes

3.1 Release Notes

Jan 31, 2026

 

This update emphasizes cost accuracy and attribution across the platform, with real billing costs now powering Recommendations and APC workloads, alongside major improvements to Workspace Cost tables, Optimization Review filters and KPIs, and Incident detection capabilities. Forecasting receives fixes for negative values and extended horizons, while dozens of chart, navigation, and performance refinements deliver a more polished day-to-day experience.

 

image-20260218-143807.png
Optimization review

Optimization Insights

Continued refinement of filters, KPIs, and default behavior in the Optimization Review experience.

  • Filter and search fixes — filters now properly reset when switching from a specific search to "All Selection"; fixed the issue where pipeline filter text persisted in the search box after reset; corrected the default Before/After date range that previously defaulted to the same month depending on user timezone

  • KPI sorting and accuracy — fixed KPI sorting not being applied after toggling between Average and Median on Jobs and Pipelines tabs; corrected Median Daily Runs before/after/diff values that were always displayed as zero; fixed wrong arrow direction and color on Median Daily Runs in Workload Activity

  • Date selection — the default date selection no longer includes today (which is not meant to be selectable)

  • Pipelines naming — fixed incorrect KPI naming on the Pipelines tab within Optimization Review

  • Jobs Cost per Data — low-value Cost per Data amounts are now properly displayed instead of being hidden

  • Cost Over Time charges — fixed double-counting of "Other" charges when a parent node is selected in Cost Over Time

 

image-20260218-144524.png

Workflows & Operations

Deeper serverless integration, trend view enhancements, and improved URL-based navigation.

  • Workflow tab URL refactoring — Workflows tab now uses unique URLs for each sub-tab (Jobs, One-Time Runs), enabling direct linking and proper browser back/forward navigation; fixed loading issues and double-click URL changes

  • All-Purpose Compute tab refactoring — APC sub-tabs (Clusters, Jobs, Notebooks, One-Time Runs) now each have a unique URL for direct linking and browser history support

  • Workflow search fix — searching workflows by name with bracket characters [] no longer returns all jobs

  • Workflow run data usage — the frequency interval for data usage on workflow runs can now be changed as expected

  • Open Workload from Incidents — the "Open Workload" button from Efficiency Incidents now works correctly for One-Time Runs

  • Pipeline updates trends — added an info panel in Pipeline Updates trends; fixed broken data points on pipeline trend charts where cost was $0

  • Navigation fixes — fixed back-button navigation between Workloads pages; resolved page failures when changing subscriptions across all Workloads pages

  • Data view filter persistence — switching from Trend view to Data view no longer loses applied filter state (all job run IDs were being shown)

 

image-20260218-144600.png

Spend Insights & Forecasting

  • Workspace Cost table redesign — updated table header design with headers and subheaders; redesigned colors and styles for better readability; added total workspace count; horizontal separator for the totals row; vertical scrollability without a fixed bottom row; fixed column editor dialog behavior

  • Workspace Cost navigation — navigate directly to per-workspace Recommendations from the Workspace Cost page; fixed date range shifts for non-UTC users during navigation; kebab menu is now visible even with long workspace names

  • Workspace Cost search — searching for a specific workspace now correctly updates the total cost row to reflect filtered results instead of showing the total for all workspaces

  • Workspace Cost export — export to CSV improvements reflecting the new interface layout; included workspace count information in cost tables

  • Databricks unit discount management — added plan-name-based Databricks unit discount configuration support with warning logs for mismatches; new interface and service support for updating discount rates; removed the "Not Applicable" Databricks unit discount label; count and track mismatch warnings during consumption loading

  • Databricks unit regression fix — fixed configured Databricks unit discounts not being applied for non-serverless workloads

  • Forecasting accuracy — extended monthly forecast horizon; fixed latest-day-of-year calculation errors; prevented negative values in both resource-based and deterministic forecasts; corrected unexpected lines appearing for past years in the Daily view; fixed wrong forecasted costs retrieval for resource-based forecast (v2)

  • Cross-page navigation — fixed subscription selection when navigating from Forecasting to Workspace Cost

  • Subscription percentage rounding — fixed incorrect rounding of selectedSubscriptionsPercentage for VMs; excluded Platform and AI costs from resource cost percentage calculations

  • Consumption reliability — fixed automatic consumption failures caused by date range exceeding the limit

 

image-20260218-145312.png

Optimization Guidance

Real billing costs, new filtering capabilities, and accuracy improvements for the Recommendations engine.

  • Real APC billing costs — Recommendations now use actual APC billing costs instead of estimates, providing more accurate savings projections

  • Cost attribution fixes — fixed doubled cost in Recommendations Runs view for midnight-spanning jobs; corrected APC cost attribution for pre-existing all-purpose clusters and long-lived clusters; fixed APC Instance cost using a 10-year fallback when app_id has no slices

  • Filter by recommendation type — the main Recommendations page and detail views now support filtering by Cost Savings, Performance, or All recommendation types

  • APC jobs disclaimer — added a disclaimer for jobs with tasks running on All-Purpose Clusters

  • VM Category — added "VM Category" column to the recommendations area in the cluster configuration table

  • VM memory flagging — VMs that are lower in memory than half of the already used memory are now flagged

  • Recommendations banner — added an informational banner on the main Recommendations screen

  • Cluster instance real cost — fixed recommendation real cost not being displayed for non-all-purpose clusters

image-20260218-145401.png

 

Incident Detection & Policies

Enhanced incident controls, new filtering, and improved policy management.

  • Enhanced cost control incidents — incidents now support a minimum dollar amount threshold in addition to percentage-based increase triggers

  • Efficiency incident filtering — Efficiency Incidents can now be filtered by resource name for faster triage (API and UI)

  • Duration incident emails — exceeded and excessive duration incidents now send email notifications

  • Default date range — the Incidents page now defaults to "Last 7 Days" for a more relevant initial view

  • Incident documentation — added documentation for minimum dollar values for incidents

  • Policy sorting — added order arrow for the policies table; fixed sorting issues for "Last Update Time" and "Created At" columns along with pagination

image-20260218-145510.png

 

Reporting & Visual Clarity

Chart accuracy, new statistical displays, and improved tooltips.

  • Trend charts median — added computed Median value to trend charts alongside existing Mean

  • Run cost charts — added Min/Avg/Max values on the run cost chart; included cost chart in the single cluster view for workflow runs trend; fixed wrong cost tooltip on single cluster trend view

  • Usage Distribution — added executor count display on usage distribution charts; implemented min/max value display; Y-axis now consistently shows up to 100% for percentage charts; fixed slow chart rescaling on window resize

  • Duration chart accuracy — fixed Min/Avg/Max values for "Duration - Execution Only" on the single cluster view

  • Pipeline trend charts — fixed broken data points from pipeline trend charts where cost was $0; fixed KPI label on multiple cluster trend view

  • Vendor Consolidation charts — fixed empty-state display in Vendor Consolidation transferred data and cost charts

  • Cost Over Time colors — updated Cloud and Databricks cost category colors for better distinction

  • Metrics unavailable banner — added a "Metrics unavailable for this date range" banner on activity histograms when data is not available

  • Serverless tables — updated serverless notebooks table, serverless pipelines table and view table, serverless job runs table, and AI Vector Search / Model Serving tables for consistency

 

image-20260218-145751.png

Navigation & Quality of Life

Smoother navigation, URL-based routing, and usability improvements.

  • Databricks navigation — updated "Navigate to Databricks" option on serverless job run view; updated Databricks navigation option on platform apps; implemented single run view navigation for serverless jobs

  • APC Notebooks table — updated APC notebooks table design; made notebook path clickable for APCs

  • Sidebar icons (iPad) — Spend Insights, Workloads, Optimize, and Settings icons now function as clickable elements on iPad Pro resolution

  • Pending labels — "Pending" cost labels are no longer shown for old runs that legitimately have no cost; added pending labels for job runs and recommendations where appropriate

  • AWS S3 tooltip — added AWS S3 tooltip guidance for Azure deployments (tips and tricks)

  • Vendor Consolidation banner — added blue documentation and guidance banner for Vendor Consolidation

  • Discount plan tooltips — embedded discount plans into tooltips for demo environments

  • SQL Warehouse config — clarified "SQL Warehouse" configuration guidance for system tables

  • Databricks unit label cleanup — removed "Databricks unit Discounts -> Not Applicable" label

 

Deployment & Setup

Reliability improvements for DBFS-disabled environments and deployment enhancements.

  • DBFS-disabled support — use workspace files when DBFS is disabled; changed volume init script naming for non-DBFS workspaces; delete global init script and update init script when DBFS is off

  • Auto Pilot metadata — exposed "last update at" and "last update by" fields for Auto Pilot configuration

  • Pipeline provisioning — temporarily disabled pipeline provisioning pending stability improvements

  • Google SSO on AWS — fixed support for Google SSO authentication in AWS Databricks deployments

  • Published workspace counts — new service returning total published and unpublished workspace counts

  • Analysis Stats — set default page size to "All" for the Analysis Stats panel

  • Configurable queue name — ANALYSIS_NOTIFICATION_QUEUE_NAME is now configurable

  • APC cluster jobs optimization — optimized the All-Purpose Cluster Jobs service for faster response times

  • APC cluster instances pagination — added pagination for APC cluster instances

  • Median aggregation — fixed median aggregation for multi-day cluster instances

  • Cluster configuration display — updated "Cluster Configuration Changes" and "Cluster Configuration" windows to display date and time