...
The
must be deployed in Microsoft Azure.D s platform Info NOTE: If you are using Azure Databricks as a datasource, please verify that openJDKv1.8.0_242 212 or earlier is installed on the
. Java 8 is required. If necessary, downgrade the Java version and restart the platform. There is a known issue with TLS v1.3.D s node
...
| Parameter | Description | Value | ||
|---|---|---|---|---|
| feature.parameterization.matchLimitOnSampling.databricksSpark | Maximum number of parameterized source files that are permitted for matching in a single dataset with parameters. | |||
| databricks.workerNodeType | Type of node to use for the Azure Databricks Workers/Executors. There are 1 or more Worker nodes per cluster. | Default:
For more information, see the sizing guide for Azure Databricks. | ||
| databricks.sparkVersion | Azure Databricks cluster version which also includes the Spark Version. | Depending on your version of Azure Databricks, please set this property according to the following:
Please do not use other values. For more information, see Configure for Spark. | ||
| databricks.serviceUrl | URL to the Azure Databricks Service where Spark jobs will be run (Example: https://westus2.azuredatabricks.net) | |||
| databricks.minWorkers | Initial number of Worker nodes in the cluster, and also the minimum number of Worker nodes that the cluster can scale down to during auto-scale-down | Minimum value: Increasing this value can increase compute costs. | ||
| databricks.maxWorkers | Maximum number of Worker nodes the cluster can create during auto scaling | Minimum value: Not less than Increasing this value can increase compute costs. | ||
| databricks.poolId | If you have enabled instance pooling in Azure Databricks, you can specify the pool identifier here. For more information, see Configure instance pooling below. |
| ||
| databricks.poolName | If you have enabled instance pooling in Azure Databricks, you can specify the pool name here. For more information, see Configure instance pooling below. | See previous.
| ||
| databricks.driverNodeType | Type of node to use for the Azure Databricks Driver. There is only 1 Driver node per cluster. | Default: For more information, see the sizing guide for Databricks.
| ||
| databricks.logsDestination | DBFS location that cluster logs will be sent to every 5 minutes | Leave this value as /trifacta/logs. | ||
| databricks.enableAutotermination | Set to true to enable auto-termination of a user cluster after N minutes of idle time, where N is the value of the autoterminationMinutes property. | Unless otherwise required, leave this value as true. | ||
| databricks.clusterStatePollerDelayInSeconds | Number of seconds to wait between polls for Azure Databricks cluster status when a cluster is starting up | |||
| databricks.clusterStartupWaitTimeInMinutes | Maximum time in minutes to wait for a Cluster to get to Running state before aborting and failing an Azure Databricks job | |||
| databricks.clusterLogSyncWaitTimeInMinutes | Maximum time in minutes to wait for a Cluster to complete syncing its logs to DBFS before giving up on pulling the cluster logs to the
| Set this to 0 to disable cluster log pulls. | ||
| databricks.clusterLogSyncPollerDelayInSeconds | Number of seconds to wait between polls for a Databricks cluster to sync its logs to DBFS after job completion | |||
| databricks.autoterminationMinutes | Idle time in minutes before a user cluster will auto-terminate. | Do not set this value to less than the cluster startup wait time value. | ||
| spark.useVendorSparkLibraries | When |
|
...
D s config method ws Locate the following property and set it to one of the values listed below:
Code Block Databricks Job Management
Property Value Description Never Delete (default) Job definitions are never deleted from the Azure Databricks cluster. Always Delete The Azure Databricks job definition is deleted during the clean-up phase, which occurs after a job completes. Delete Successful Only When a job completes successfully, the Azure Databricks job definition is deleted during the clean-up phase. Failed or canceled jobs are not deleted, which allows you to debug as needed. Skip Job Creation For jobs that are to be executed only one time, the
can be configured to use a different mechanism for submitting the job. When this option is enabled, theD s platform
submits jobs using the run-submit API, instead of the run-now API. The run-submit API does not create an Azure Databricks job definition. Therefore the submitted job does not count toward the enforced job limit.D s platform Default Inherits the default system-wide setting. - Save your changes and restart the platform.
...
To enable, the following parameters for the
| D s platform |
|---|
| D s config |
|---|
| Parameter | Description |
|---|---|
| azure.managedIdentities.enabled | Set to true to enable use of Azure managed identities. |
| azure.managedIdentities.keyVaultApplicationidSecretName | Specify the name of the Azure Key Vault secret that holds the service principal Application Id. |
| azure.managedIdentities.keyVaultApplicationSecretSecretName | Specify the name of the Key Vault secret that holds the service principal secret. |
Save your changes.
Pass additional Spark properties
...
D s config - Acquire the value for
databricks.sparkVersion. In Azure Databricks, compare your value to the list of supported Azure Databricks version. If your version is unsupported, identify a new version to use.
Info NOTE: Please make note of the version of Spark supported for the version of Azure Databricks that you have chosen.
- In the
configuration:D s platform - Set
databricks.sparkVersionto the new version to use. - Set
spark.versionto the appropriate version of Spark to use.
- Set
- Restart the
.D s platform - The platform is restarted. A new Azure Databricks cluster is created for each user using the specified values, when the user runs a job.
Spark job fails with "spark scheduler cannot be cast" error
When you run a job on Databricks, the job may fail with the following error:
| Code Block |
|---|
java.lang.ClassCastException: org.apache.spark.scheduler.ResultTask cannot be cast to org.apache.spark.scheduler.Task
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:616)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748) |
The job.log file may contain something similar to the following:
| Code Block |
|---|
2022-07-19T15:41:24.832Z - [sid=0cf0cff5-2729-4742-a7b9-4607ca287a98] - [rid=83eb9826-fc3b-4359-8e8f-7fbf77300878] - [Async-Task-9] INFO com.trifacta.databricks.spark.JobHelper - Got error org.apache.spark.SparkException: Stage 0 failed. Error: Job aborted due to stage failure: Task 0 in stage 0.0 failed 4 times, most recent failure: Lost task 0.3 in stage 0.0 (TID 3, ip-10-243-149-238.eu-west-1.compute.internal, executor driver): java.lang.ClassCastException: org.apache.spark.scheduler.ResultTask cannot be cast to org.apache.spark.scheduler.Task
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:616)
... |
This error is due to a class mismatch between the
| D s platform |
|---|
Solution:
The solution is to disable the precedence of using the Spark JARs provided from the
| D s platform |
|---|
D s config method t Locate the
spark.propssection and add the following configuration elements:Code Block "spark": { ... "props": { "spark.driver.userClassPathFirst": false, "spark.executor.userClassPathFirst": false, ... } },- Save your changes and restart the platform.
