Documentation Home
MySQL NDB Cluster 8.3
Related Documentation Download this Excerpt
PDF (US Ltr) - 4.6Mb
PDF (A4) - 4.6Mb

MySQL NDB Cluster 8.3  /  Management of NDB Cluster  /  NDB Cluster and the Performance Schema

6.19 NDB Cluster and the Performance Schema

NDB provides information in the MySQL Performance Schema about ndbcluster plugin threads and instrumenting for transaction batch memory. These features are described in greater detail in the sections which follow.

ndbcluster Plugin Threads

ndbcluster plugin threads are visible in the Performance Schema threads table, as shown in the following query:

mysql> SELECT name, type, thread_id, thread_os_id
    -> FROM performance_schema.threads
    -> WHERE name LIKE '%ndbcluster%'\G
| name                             | type       | thread_id | thread_os_id |
| thread/ndbcluster/ndb_binlog     | BACKGROUND |        30 |        11980 |
| thread/ndbcluster/ndb_index_stat | BACKGROUND |        31 |        11981 |
| thread/ndbcluster/ndb_metadata   | BACKGROUND |        32 |        11982 |

The threads table shows all three of the threads listed here:

  • ndb_binlog: Binary logging thread

  • ndb_index_stat: Index statistics thread

  • ndb_metadata: Metadata thread

These threads are also shown by name in the setup_threads table.

Thread names are shown in the name column of the threads and setup_threads tables using the format prefix/plugin_name/thread_name. prefix, the object type as determined by the performance_schema engine, is thread for plugin threads (see Thread Instrument Elements). The plugin_name is ndbcluster. thread_name is the standalone name of the thread (ndb_binlog, ndb_index_stat, or ndb_metadata).

Using the thread ID or OS thread ID for a given thread in the threads or setup_threads table, it is possible to obtain considerable information from Performance Schema about plugin execution and resource usage. This example shows how to obtain the amount of memory allocated by the threads created by the ndbcluster plugin from the mem_root arena by joining the threads and memory_summary_by_thread_by_event_name tables:

mysql> SELECT
    ->   m.sum_number_of_bytes_alloc,
    ->   IF(m.sum_number_of_bytes_alloc > 0, "true", "false") AS 'Has allocated memory'
    -> FROM performance_schema.memory_summary_by_thread_by_event_name m
    -> JOIN performance_schema.threads t
    -> ON m.thread_id = t.thread_id
    -> WHERE LIKE '%ndbcluster%'
    ->   AND event_name LIKE '%THD::main_mem_root%';
| name                             | sum_number_of_bytes_alloc | Has allocated memory |
| thread/ndbcluster/ndb_binlog     |                     20576 | true                 |
| thread/ndbcluster/ndb_index_stat |                         0 | false                |
| thread/ndbcluster/ndb_metadata   |                      8240 | true                 |

Transaction Memory Usage

You can see the amount of memory used for transaction batching by querying the Performance Schema memory_summary_by_thread_by_event_name table, similar to what is shown here:

    ->   FROM performance_schema.memory_summary_by_thread_by_event_name
    ->     AND EVENT_NAME LIKE 'memory/ndbcluster/%';
| EVENT_NAME                                |
| memory/ndbcluster/Thd_ndb::batch_mem_root |
1 row in set (0.01 sec)

The ndbcluster transaction memory instrument is also visible in the Performance Schema setup_instruments table, as shown here:

mysql> SELECT * from performance_schema.setup_instruments
    ->   WHERE NAME LIKE '%ndb%'\G
*************************** 1. row ***************************
         NAME: memory/ndbcluster/Thd_ndb::batch_mem_root
        TIMED: NULL
DOCUMENTATION: Memory used for transaction batching
1 row in set (0.01 sec)