4.7 Release Notes

What's New

Engine Tags

Dremio administrators can add engine tags for all AWS EC2 objects associated with each engine in an AWS Edition deployment. Add the tags on the Edit Elastic Engine modal in the Dremio user interface.

Note:
Dremio users cannot add engine tags when initially deploying Dremio AWS Edition because engines are created dynamically.

Arrow Caching

Dremio users can improve query performance by enabling Arrow caching, which caches Data Reflections in the Apache Arrow format. Users enable the feature for each reflection by clicking the Switch to Advanced button on the Dataset Settings modal, then toggling Arrow caching to the on position on the Settings: Raw Reflection and Settings: Aggregation Reflection modals.

Secondary Coordinator Nodes

Dremio provides improved concurrency and distribution of query loads across a deployment with secondary coordinator nodes. To enable this feature for YARN-based and standalone deployments, add the following to the dremio.conf file on each secondary coordinator node in your cluster and restart your deployment:

services: {
    coordinator.enabled: true,
    coordinator.master.enabled: false,
    coordinator.web.enabled: false,
    executor.enabled: false
}

To enable secondary coordinator nodes for Azure AKE and AWS EKS deployments, update coordinator.count in values.yaml of the helm chart and restart your deployment.

Tip
Dremio recommends a maximum of five secondary coordinator nodes for a deployment.

Note:
Dremio deployments on AWS Edition and Azure ARM do not support secondary coordinator nodes.

Kubernetes Helm Charts

Dremio 4.7.0 features an updated Helm chart for Kubernetes-based deployments using Helm 3.0.0+. The updated chart no longer supports Helm 2.x.

Preview: Runtime Filtering

Dremio optimizes the performance of JOIN statements by dynamically applying filters from dimension tables to fact tables. Contact Dremio Customer Support to enable runtime filtering.

Disabling Cross Source Queries

Dremio administrators can disable single queries that select data across multiple data sources, including virtual and physical datasets. To disable cross-source SELECT statements, administrators set the planner.cross_source_select.disable support key to true. By default, the feature is disabled.

Note
Dremio users can override this feature by checking Enable this soure to be used even though Disable Cross Source is configured on the Advanced Options tab of the Edit Source modal.

Separation of Data Lake and External Data Sources

The Dremio Datasets page now separates data sources into two categories:

  • Data Lakes - filesystem and table-based data sources
  • External Sources - RDBMS and NoSQL-based data sources

Changed Behavior

  • Dremio maps the DATE type in Oracle data sources to the TIMESTAMP Dremio data type. Prior to version 4.7.0, Dremio mapped DATE to the Dremio DATE data type. Dremio developers using Oracle data sources must update their application logic before running their applications against a Dremio deployment upgraded to Dremio 4.7.0. Developers must update their applicaton code to expect a TIMESTAMP data type rather than a DATE type.

  • Kubernetes-based deployments no longer support Helm 2.x. Dremio 4.7.0 features an updated chart that requires Helm 3.0.0+.

  • Dremio disables the Edit Settings button on the Engines page when starting and deleting an engine.

Fixed Issues in 4.7.0

Selected table has no columns error when querying INFORMATION_SCHEMA.COLUMNS
Fixed by detecting missing fields in BatchSchema entries for tables.

TIMESTAMPDIFF() function returns different values with the same inputs for optimized queries
Fixed issue with timestampdiffMonth() function when the end date is the last day of a month.

NamespaceNotFoundException: one or more elements on the path are not found in namespace
Fixed by exploring dataset rather than parsing resourceId.

The catalog API fails to remove artifacts after a a failed PUT request to the /api/v3/catalog REST API
Updated the validation of AWS S3 configuration paths.

Query returns AssertionError: Type mismatch
Fixed mismatched type during fragment initialization.

/api/v3/catalog/<id>/graph REST API returns NullPointerException
Updated the logic that addresses data sources for virtual datasets.

SQL Editor displays Failed to parse date time value <timestamp> in field timestamp error
Fixed bug that generated child field paths with an incorrect parent path.

Tableau Server unable to fetch data with ODBC/JDBC connection with Synchronous socket write failed with error: An established connection was aborted by the software in your host machine error
Fixed by updating the logic that maps RpcException to SQLStateCode in the JDBC driver.

Unable to relaunch Dremio AWS Edition with CloudFormation due to permissions error
Removed hard-coded IAM role and security group to launch Dremio AWS Edition with CloudFormation.

Inconsistent results when using an OR clause while querying INFORMATION_SCHEMA
Fixed an incorrect expression filter push down in queries with an OR clause.

Dremio user interface crashes when editing decimal fields
Fixed bug in handling decimals in the bar chart.

Queries against Postgres datasets fail with ERROR: window functions are not allowed in window definitions error
Fixed bug that created a nested window in RelToSqlConverter.

Performance issues when a Dremio user repeatedly queries the same Teradata table
Fixed bug that calculated the row count twice for an RDBMS plugin.

Inconsistent results when querying virtual datasets
Fixed a race condition during node completion.

UNAVAILABLE: Channel shutdown invoked error
Fixed bug where jobclient re-caches the channel.

4.7.1 Release Notes

Fixed Issues in 4.7.1

Executor node crashes during query
Moved execution of casting to the INT and BIGINT data types with the CAST SQL function to Java rather than Gandiva.

Dremio fails to deduplicate project expressions
Added code to ensure Dremio deduplicates prewarmed expressions correctly.

4.7.2 Release Notes

Fixed Issues in 4.7.2

Text overflows in the Datasets navigation pane in Firefox and Safari
Fixed by using an overflow property to hide out-of-view spaces and data sources.

4.7.3 Release Notes

What's New

External Queries

External queries enable Dremio users to query relational database data sources using the SQL syntax native to that database. Users can save the results of their external queries as Dremio virtual datasets and enable Data Reflections.

Fixed Issues in 4.7.3

Queries fail on MongoDB data sources when Dremio pushes down "IS NULL" and "IS NOT NULL
Optimized pushdown for MongoDB by using a comparison expression rather than an aggregate expression and fixed a bug that returned a superfluous resultset.

Unable to specify a storage class in the Helm chart
Fixed a typo that prevented custom storage classes from being used in Dremio Helm Chart v2.

Dremio returns inconsistent results for vectorized JOIN SQL statements when Runtime Filtering is enabled
Set the validity bit from right for VECTORIZED_BIGINT mode.


results matching ""

    No results matching ""