Inheritance of index and source declarations

Inheritance in index and source declarations enables better organization of tables with similar settings or structures and reduces the configuration size. Both parent and child tables or sources can utilize inheritance.

No specific configurations are needed for a parent table or source.

In the child table or source declaration, specify the table or source name followed by a colon (:) and the parent name:

table parent {
path = /var/lib/manticore/parent
...
}

table child:parent {
path = /var/lib/manticore/child
...
}

The child will inherit the entire configuration of the parent. Any settings declared in the child will overwrite the inherited values. Be aware that for multi-value settings, defining a single value in the child will clear all inherited values. For example, if the parent has several sql_query_pre declarations and the child has a single sql_query_pre declaration, all inherited sql_query_pre declarations are cleared. To override some of the inherited values from the parent, explicitly declare them in the child. This is also applicable if you don't need a value from the parent. For example, if the sql_query_pre value from the parent is not needed, declare the directive with an empty value in the child like sql_query_pre=.

Note that existing values of a multi-value setting will not be copied if the child declares one value for that setting.

The inheritance behavior applies to fields and attributes, not just table options. For example, if the parent has two integer attributes and the child needs a new integer attribute, the integer attribute declarations from the parent must be copied into the child configuration.

Setting variables online

SET

SET [GLOBAL] server_variable_name = value
SET [INDEX table_name] GLOBAL @user_variable_name = (int_val1 [, int_val2, ...])
SET NAMES value [COLLATE value]
SET @@dummy_variable = ignored_value

The SET statement in Manticore Search allows you to modify variable values. Variable names are case-insensitive, and no variable value changes will persist after a server restart.

Manticore Search supports the SET NAMES statement and SET @@variable_name syntax for compatibility with third-party MySQL client libraries, connectors, and frameworks that may require running these statements when connecting. However, these statements do not have any effect on Manticore Search itself.

There are four classes of variables in Manticore Search:

  1. Per-session server variable: set var_name = value
  2. Global server variable: set global var_name = value
  3. Global user variable: set global @var_name = (value)
  4. Global distributed variable: set index dist_table_name global @var_name = (value)

Global user variables are shared between concurrent sessions. The only supported value type is a list of BIGINTs, and these variables can be used with the IN() operator for filtering purposes. The primary use case for this feature is to upload large lists of values to searchd once and reuse them multiple times later, reducing network overhead. Global user variables can be transferred to all agents of a distributed table or set locally in the case of a local table defined in a distributed table. Example:

// in session 1
mysql> SET GLOBAL @myfilter=(2,3,5,7,11,13);
Query OK, 0 rows affected (0.00 sec)

// later in session 2
mysql> SELECT * FROM test1 WHERE group_id IN @myfilter;
+------+--------+----------+------------+-----------------+------+
| id   | weight | group_id | date_added | title           | tag  |
+------+--------+----------+------------+-----------------+------+
|    3 |      1 |        2 | 1299338153 | another doc     | 15   |
|    4 |      1 |        2 | 1299338153 | doc number four | 7,40 |
+------+--------+----------+------------+-----------------+------+
2 rows in set (0.02 sec)

Manticore Search supports per-session and global server variables that affect specific server settings in their respective scopes. Below is a list of known per-session and global server variables:

Known per-session server variables:

  • AUTOCOMMIT = {0 | 1} determines if data modification statements should be implicitly wrapped by BEGIN and COMMIT.

  • COLLATION_CONNECTION = collation_name selects the collation for ORDER BY or GROUP BY on string values in subsequent queries. Refer to Collations for a list of known collation names.

  • MAX_THREADS_PER_QUERY = <POSITIVE_INT_VALUE> redefines max_threads_per_query in the runtime. Per-session variable influences only the queries run in the same session (connection), i.e., up to disconnect. Value 0 means 'no limit'. If both per-session and the global variables are set, the per-session one has a higher priority.

  • net_write_timeout = <value>: Tunes the network timeout for write operations, i.e., sending data. The global value can be changed only with VIP privileges.

  • optimize_by_id = {0 | 1}: Internal flag used in some debug commands.

  • PROFILING = {0 | 1} enables query profiling in the current session. Defaults to 0. See also show profile.

  • ro = {1 | 0} switches session to read-only mode or back. In show variables output the variable displayed with name session_read_only.

  • throttling_period = <INT_VALUE>: Interval (in milliseconds) during which the current running query will reschedule. A value of 0 disables throttling, meaning the query will occupy CPU cores until it finishes. If concurrent queries come from other connections at the same time, they will be allocated to free cores or will be suspended until a core is released. Providing a negative value (-1) resets throttling to the default compiled-in value (100ms), which means the query will be rescheduled every 100ms, allowing concurrent queries a chance to be executed. The global value (set via set global) can only be set on a VIP connection.

  • thread_stack = <value>: Changes the default value on-the-fly, which limits the stack size provided to one task. Note that here 'thread' refers not to an OS thread, but to a userspace thread, also known as a coroutine. This can be useful if, for example, you load a percolate table with unexpectedly high requirements. In such cases, 'call pq' would fail with a message about insufficient stack size. Generally, you should stop the daemon, increase the value in the config, and then restart. However, you can also try a new value without restarting, by setting a new one with this variable. The global value can also be changed online with set global thread_stack, but this is available only from a VIP connection.

  • threads_ex (diagnostic): Forces Manticore to behave as if it is running on a CPU with the provided profile. As a short example, set threads_ex='4/2+6/3' indicates 'you have 4 free CPU cores, when scheduling multiple queries they should be batched by 2. Also, you have 6 free CPU cores for pseudo-sharding, parts should be batched by 3'. This option is diagnostic, as it is very helpful, for example, to see how your query would run on a configuration you don't have locally. For instance, on a 128-core CPU. Or, conversely, to quickly limit the daemon to behave as single-threaded, to locate a bottleneck or investigate a crash.

    MORE

  • WAIT_TIMEOUT/net_read_timeout = <value> sets connection timeout, either per session or global. Global can only be set on a VIP connection.

Known global server variables are:

  • ACCURATE_AGGREGATION: Sets the default value for the option accurate_aggregation of future queries.
  • AUTO_OPTIMIZE = {1|0} Turns on/off auto_optimize.
  • cluster_user = name Sets the username used with mysqldump / mariadb-dump to enable replication mode.
  • COREDUMP= {1|0} Turns on/off saving a core file or a minidump of the server on crash. More details here.
  • CPUSTATS= {1|0} Turns on/off CPU time tracking.
  • DISTINCT_PRECISION_THRESHOLD: Sets the default value for the option distinct_precision_threshold of future queries.
  • ES_COMPAT = {on/off/dashboards} When set to on (default), Elasticsearch-like write requests are supported; off disables the support; dashboards enables the support and also allows requests from Kibana (this functionality is experimental).
  • EXPANSION_MERGE_THRESHOLD_DOCS: Changes the value of the config's expansion_merge_threshold_docs setting on-the-fly.
  • EXPANSION_MERGE_THRESHOLD_HITS: Changes the value of the config's expansion_merge_threshold_hits setting on-the-fly.
  • GROUPING_IN_UTC = {0 | 1} When set to 1, causes timed grouping functions (day(), month(), year(), yearmonth(), yearmonthday()) to be calculated in UTC. Read the doc for grouping_in_utc config params for more details.
  • IOSTATS = {0 | 1} Enables or disables I/O operations (except for attributes) reporting in the query log.
  • LOG_DEBUG_FILTER = <string value> Filters out redundant log messages. If the value is set, then all logs with level > INFO (i.e., DEBUG, DEBUGV, etc.) will be compared with the string and output only in the case they start with the given value.
  • LOG_LEVEL = {info | debug | replication | debugv | debugvv} Changes the current log verboseness level.
  • MAINTENANCE = {0 | 1} When set to 1, puts the server in maintenance mode. Only clients with VIP connections can execute queries in this mode. All new non-VIP incoming connections are refused. Existing connections are left intact.
  • MAX_THREADS_PER_QUERY = <POSITIVE_INT_VALUE> Redefines max_threads_per_query at runtime. As global, it changes behavior for all sessions. Value 0 means 'no limit'. If both per-session and global variables are set, the per-session one has a higher priority.
  • NET_WAIT = {-1 | 0 | POSITIVE_INT_VALUE} Changes the net_wait_tm searchd settings value.
  • OPTIMIZE_CUTOFF = <value>: Changes the value of the config's optimize_cutoff setting on-the-fly.
  • PSEUDO_SHARDING = {1|0} Turns on/off search pseudo-sharding.
  • QCACHE_MAX_BYTES = <value> Changes the query_cache RAM use limit to a given value.
  • QCACHE_THRESH_MSEC = <value> Changes the query_cache minimum wall time threshold to a given value.
  • QCACHE_TTL_SEC = <value> Changes the query_cache TTL for a cached result to a given value.
  • QUERY_LOG_FORMAT = {plain | sphinxql} Changes the current log format.
  • QUERY_LOG_MIN_MSEC = <value> Changes the query_log_min_msec searchd settings value. In this case, it expects the value exactly in milliseconds and doesn't parse time suffixes, as in config.

    Warning: this is a very specific and 'hard' variable; filtered out messages will be just dropped and not written into the log at all. Better just filter your log with something like 'grep', in this case, you'll have at least the full original log as a backup.

  • RESET_NETWORK_TIMEOUT_ON_PACKET = {1|0} changes reset_network_timeout_on_packet param. Only clients with VIP connections can change this variable.
  • SECONDARY_INDEXES = {1|0} Turns on/off secondary indexes for search queries.
  • TIMEZONE = <value> Specifies the timezone used by date/time-related functions. Read the doc for timezone config param for more details.

Examples:

mysql> SET autocommit=0;
Query OK, 0 rows affected (0.00 sec)

mysql> SET GLOBAL query_log_format=sphinxql;
Query OK, 0 rows affected (0.00 sec)

mysql> SET GLOBAL @banned=(1,2,3);
Query OK, 0 rows affected (0.01 sec)

mysql> SET INDEX users GLOBAL @banned=(1,2,3);
Query OK, 0 rows affected (0.01 sec)

To make user variables persistent, make sure sphinxql_state is enabled.

⪢ Integration