7.3 HeatWave AutoML Monitoring

You can monitor HeatWave AutoML status by querying the rapid_ml_status variable or by querying the ML_STATUS column of the performance_schema.rpd_nodes table.

• Querying the rapid_ml_status variable:

  The rapid_ml_status variable provides the status of HeatWave AutoML. Possible values are ON and OFF.

  • ON: HeatWave AutoML is up and running.

  • OFF: HeatWave AutoML is down.

  You can query the rapid_ml_status status variable directly or through the performance_schema.global_status table; for example:

  mysql> SHOW GLOBAL STATUS LIKE 'rapid_ml_status';
  +-----------------+-------+
  | Variable_name   | Value |
  +-----------------+-------+
  | rapid_ml_status | ON    |
  +-----------------+-------+

  mysql> SELECT VARIABLE_NAME, VARIABLE_VALUE
            FROM performance_schema.global_status
            WHERE VARIABLE_NAME LIKE 'rapid_ml_status';
  +-----------------+----------------+
  | VARIABLE_NAME   | VARIABLE_VALUE |
  +-----------------+----------------+
  | rapid_ml_status | ON             |
  +-----------------+----------------+

• Querying the ML_STATUS column of the performance_schema.rpd_nodes table.

  The HeatWave plugin writes HeatWave AutoML status information to the ML_STATUS column of performance_schema.rpd_nodes table after each ML query. Possible values include:

  • UNAVAIL_MLSTATE: HeatWave AutoML is not available.

  • AVAIL_MLSTATE: HeatWave AutoML is available.

  • DOWN_MLSTATE: HeatWave AutoML is down.

  ML_STATUS is reported for each HeatWave node.

  To following query retrieves ID, STATUS, and ML_STATUS for each HeatWave node from the performance_schema.rpd_nodes table:

  mysql> SELECT ID, STATUS, ML_STATUS FROM performance_schema.rpd_nodes; 
  +----+---------------+---------------+
  | ID | STATUS        | ML_STATUS     |
  +----+---------------+---------------+
  |  1 | AVAIL_RNSTATE | AVAIL_MLSTATE |
  |  0 | AVAIL_RNSTATE | AVAIL_MLSTATE |
  +----+---------------+---------------+

If rapid_ml_status is OFF or ML_STATUS reports DOWN_MLSTATE for any HeatWave node, you can restart the HeatWave Cluster in the HeatWave Console but be aware that restarting interrupts any analytics queries that are running. See Managing a HeatWave Cluster in the HeatWave on OCI Service Guide.