Job Manager

The Job Manager maintains request ordering across the nodes in your Kinetica cluster at the cost of some performance. All read requests (/show/table, /get/records, etc.) are processed simultaneously. When a write request (/insert/records, /delete/records, etc.) is made, all subsequent read and write requests are blocked until the current write request is finished.

If Job Manager is disabled, all inbound requests are broadcasted to all worker nodes and can be executed in any order.

While enabled by default, Job Manager can be disabled by setting the enable_job_manager configuration parameter to false in the gpudb.conf file.

The Job Manager interface is available through GAdmin, which allows jobs to be monitored and cancelled, if needed.

Because read requests are processed simultaneously, read jobs can only be cancelled while a preceding write request is being processed, which is also referred to as a checkpoint.

Important

Insert, update, and delete requests cannot be cancelled. Otherwise, most user-facing non-admin jobs are cancellable.

When using any endpoint, you can specify a name, number, or other tag to help identify a particular job using the job_user_data options parameter; e.g., (in Python):

1
2
3
4
5
6
7
h_db.insert_records(
  table_name = "test_table_all_types",
  data = encoded_obj_list, list_encoding = 'binary',
  options = {
    "job_user_data":"insert_records_large_table"
  }
)