MySQL AI  /  ...  /  The rpd_nodes Table

6.1.4 The rpd_nodes Table

The rpd_nodes table provides information about AI nodes.

The rpd_nodes table has these columns:

  • ID

    A unique identifier for the MySQL AI Engine (AI engine).

  • CORES

    The number of cores used by the AI engine.

  • MEMORY_USAGE

    Node memory usage in bytes. The value is refreshed every four seconds. If a query starts and finishes in the four seconds between refreshes, the memory used by the query is not accounted for in the reported value.

  • MEMORY_TOTAL

    The total memory in bytes allocated to the AI engine.

  • BASEREL_MEMORY_USAGE

    The base relation memory footprint per node.

  • STATUS

    The status of the AI engine. Possible statuses include:

    • NOTAVAIL_RNSTATE

      Not available.

    • AVAIL_RNSTATE

      Available.

    • DOWN_RNSTATE

      Down.

    • DEAD_RNSTATE

      The node is not operational.

  • IP

    IP address of the AI engine.

  • PORT

    The port on which the AI engine was started.

  • CLUSTER_EVENT_NUM

    The number of cluster events such as node down, node up, and so on.

  • NUM_OBJSTORE_GETS

    Number of GET requests from the AI engine to the disk.

  • NUM_OBJSTORE_PUTS

    The number of PUT requests from the AI engine to the disk.

  • NUM_OBJSTORE_DELETES

    The number of DELETE requests from the AI engine to the disk.

  • ML_STATUS

    AutoML status. Possible status values include:

    • UNAVAIL_MLSTATE: AutoML is not available.

    • AVAIL_MLSTATE: AutoML is available.

    • DOWN_MLSTATE: AutoML declares the node is down.

The rpd_nodes table is read-only.

The rpd_nodes table may not show the current status for a new node or newly configured node immediately. The rpd_nodes table is updated after the node has successfully joined the cluster.

If additional nodes fail while node recovery is in progress, the newly failed nodes are not detected and their status is not updated in the performance_schema.rpd_nodes table until after the current recovery operation finishes and the nodes that failed previously have rejoined the cluster.