Page tree

Release 8.2.1



Contents:

   

Contents:


The Trifacta® platform utilizes the batch job runner service to orchestrate jobs that are executed on the selected backend running environment. This service passes jobs to the backend and tracks their progress until success or failure. This service is enabled by default.

Configure Timeout

SettingDefault ValueDescription
batchserver.spark.requestTimeoutMillis120000 (2 minutes)Maximum number of milliseconds that the Batch Job Runner service should wait for a response from the Spark Job service during job execution. Default is 2 minutes.

Configure polling intervals

The following parameters can be modified to change the Batch Job Runner polling intervals for various types of jobs. 

SettingDefault ValueDescription
jobMonitoring.ingestPollFrequencySeconds3Polling interval in seconds for Batch Job Runner to check for status of ingest jobs.
jobMonitoring.publishPollFrequencySeconds3Polling interval in seconds for Batch Job Runner to check for status of publishing jobs.
jobMonitoring.wranglePollFrequencySeconds3Polling interval in seconds for Batch Job Runner to check for status of wrangling jobs.
jobMonitoring.maxHeartbeatDelayMilliseconds7200000 (2 hours)Duration in milliseconds for the service to wait for a job heartbeat before failing.

Configure Job Threads

As needed, you can configure the number of worker threads assigned to each process that is managed by the batch job runner. Depending of the volume and complexity of jobs that you run of each type, you may choose to modify these settings to improve performance for key job types. 

Tip: These settings can be configured through the Admin Settings page in the Trifacta application. See Admin Settings Page.


By running environment:

SettingDefault ValueDescription
batchserver.workers.photon.max2

Number of worker threads for running Trifacta Photon jobs. This value corresponds to the maximum number of photon jobs that can be queued at the same time.

For more information, see Configure Photon Running Environment.

batchserver.workers.spark.max16Number of worker threads for running Spark jobs. For more information, see Configure Spark Running Environment.
batchserver.workers.wrangle.max16

Number of worker threads for running transformation jobs.


By job type:

SettingDefault ValueDescription
batchserver.workers.ingest.max
16Maximum number of worker threads for running ingest jobs, which are used for loading relational data into the platform. After this maximum number has been reached, subsequent requests are queued.
batchserver.workers.profile.max
16 Maximum number of worker threads for running profile jobs, which provide summary and detail statistics on job results. 
batchserver.workers.publish.max
16 Maximum number of worker threads for running publish jobs, which deliver pre-generated job results to other datastores. 
batchserver.workers.fileconverter.max
16 Maximum number of worker threads for running fileconverter jobs, which are used to convert source formats into output formats. 
batchserver.workers.filewriter.max
16 Maximum number of worker threads for running filewriter jobs, which are used for writing file-based outputs to a specified storage location.

 

Depending on your running environment, there may be additional parameters that you can configure to affect Batch Job Runner for that specific environment:

Configure BJR for EMR

Multiple BJR instances

If the Trifacta platform is connected to an EMR cluster, multiple instances of the batch job runner are deployed to manage jobs across the cluster so that if one fails, YARN jobs are still tracked. No configuration is required.

YARN logs from EMR

The following properties below can be modified for batch job runner:

SettingDefault ValueDescription
aws.emr.getLogsOnFailurefalse

When set to true, YARN logs from all nodes in the EMR cluster are collected from S3 and stored on the Trifacta node in the following location:

/opt/trifacta/logs/jobs/<jobId>/container

where: <jobId> is the Trifacta platform internal identifier for the job that failed.

aws.emr.getLogsForAllJobsfalse

When set to true, YARN logs from nodes in the EMR cluster are collected and stored in the above location for all jobs, whether they succeed or fail.

NOTE: This parameter is intended for debugging purposes only.

Configure Database

Configure database cleanup

By default, the Jobs database, which is used by the batch job runner, does not remove information about jobs after they have been executed. 

Logging:

  • Batch Job Runner activities are surfaced in batch-job-runner.log. For more information, see Configure Logging for Services.
  • Logging information for individual jobs is available in the job.log file written in the job directory. For more information, see Diagnose Failed Jobs.

As needed, you can enable the Trifacta platform to perform cleanup operations on the Jobs database.

NOTE: If cleanup is not enabled, the Jobs database continues to grow. You should perform periodic cleanups in conjunction with your enterprise database policies.

Steps:

To enable cleanup of the Jobs database, please complete the following steps.

  1. You can apply this change through the Admin Settings Page (recommended) or
    trifacta-conf.json
    . For more information, see Platform Configuration Methods.
  2. Locate the following settings and set them accordingly:

    SettingsDescription
    batch-job-runner.cleanup.enabledSet this value to true to enable this feature.
    batch-job-runner.cleanup.intervalThe interval in ISO-8601 repeating intervals format at which batch-job-runner should clean outdated information about jobs. Default value is R/PT1H, which means that the job is executed once per hour.
    batch-job-runner.cleanup.maxAgeThe retention time for deployments in ISO-8601 interval format after which information about jobs is considered outdated. Default value is P14D, which means that the jobs information is cleaned out after 14 days.
    batch-job-runner.cleanup.maxDeleteMaximum number of jobs whose information can be deleted per cleanup pass. Default value is 1000.

    For more information on ISO-8601 interval format, see https://en.wikipedia.org/wiki/ISO_8601#Repeating_intervals.

  3. Save your changes and restart the platform.

Configure Jobs database 

The Batch Job Runner utilizes its own Jobs database. For more information, see Configure the Databases.

Logging

For more information on logging for the service, see Configure Logging for Services.

This page has no comments.