The below settings are to be used in the searchd
section of the Manticore Search configuration file to control the server's behavior. Below is a summary of each setting:
This setting sets instance-wide defaults for access_plain_attrs. It is optional, with a default value of mmap_preread
.
The access_plain_attrs
directive allows you to define the default value of access_plain_attrs for all tables managed by this searchd instance. Per-table directives have higher priority and will override this instance-wide default, providing more fine-grained control.
This setting sets instance-wide defaults for access_blob_attrs. It is optional, with a default value of mmap_preread
.
The access_blob_attrs
directive allows you to define the default value of access_blob_attrs for all tables managed by this searchd instance. Per-table directives have higher priority and will override this instance-wide default, providing more fine-grained control.
This setting sets instance-wide defaults for access_doclists. It is optional, with a default value of file
.
The access_doclists
directive allows you to define the default value of access_doclists for all tables managed by this searchd instance. Per-table directives have higher priority and will override this instance-wide default, providing more fine-grained control.
This setting sets instance-wide defaults for access_hitlists. It is optional, with a default value of file
.
The access_hitlists
directive allows you to define the default value of access_hitlists for all tables managed by this searchd instance. Per-table directives have higher priority and will override this instance-wide default, providing more fine-grained control.
This setting sets instance-wide defaults for access_dict. It is optional, with a default value of mmap_preread
.
The access_dict
directive allows you to define the default value of access_dict for all tables managed by this searchd instance. Per-table directives have higher priority and will override this instance-wide default, providing more fine-grained control.
This setting sets instance-wide defaults for the agent_connect_timeout parameter.
This setting sets instance-wide defaults for the agent_query_timeout parameter. It can be overridden on a per-query basis using the OPTION agent_query_timeout=XXX
clause.
This setting is an integer that specifies how many times Manticore will attempt to connect and query remote agents through a distributed table before reporting a fatal query error. The default value is 0 (i.e., no retries). You can also set this value on a per-query basis using the OPTION retry_count=XXX
clause. If a per-query option is provided, it will override the value specified in the configuration.
Note that if you use agent mirrors in the definition of your distributed table, the server will select a different mirror for each connection attempt according to the chosen ha_strategy. In this case, the agent_retry_count
will be aggregated for all mirrors in a set.
For example, if you have 10 mirrors and set agent_retry_count=5
, the server will retry up to 50 times, assuming an average of 5 tries for each of the 10 mirrors (with the ha_strategy = roundrobin
option, this will be the case).
However, the value provided as the retry_count
option for the agent serves as an absolute limit. In other words, the [retry_count=2]
option in the agent definition always means a maximum of 2 attempts, regardless of whether you have specified 1 or 10 mirrors for the agent.
This setting is an integer in milliseconds (or special_suffixes) that specifies the delay before Manticore retries querying a remote agent in case of failure. This value is only relevant when a non-zero agent_retry_count or non-zero per-query retry_count
is specified. The default value is 500. You can also set this value on a per-query basis using the OPTION retry_delay=XXX
clause. If a per-query option is provided, it will override the value specified in the configuration.
When using Update to modify document attributes in real-time, the changes are first written to an in-memory copy of the attributes. These updates occur in a memory-mapped file, meaning the OS decides when to write the changes to disk. Upon normal shutdown of searchd
(triggered by a SIGTERM
signal), all changes are forced to be written to disk.
You can also instruct searchd
to periodically write these changes back to disk to prevent data loss. The interval between these flushes is determined by attr_flush_period
, specified in seconds (or special_suffixes).
By default, the value is 0, which disables periodic flushing. However, flushing will still occur during a normal shutdown.
- Example
attr_flush_period = 900 # persist updates to disk every 15 minutes
This setting controls the automatic OPTIMIZE process for table compaction.
By default table compaction occurs automatically. You can modify this behavior with the auto_optimize
setting:
- 0 to disable automatic table compaction (you can still call
OPTIMIZE
manually) - 1 to explicitly enable it
- N to enable it, but allow OPTIMIZE to start when the number of disk chunks is greater than
# of CPU cores * 2 * N
Note that toggling auto_optimize
on or off doesn't prevent you from running OPTIMIZE TABLE manually.
- Disable
- Throttle
auto_optimize = 0 # disable automatic OPTIMIZE
Manticore supports the automatic creation of tables that don't yet exist but are specified in INSERT statements. This feature is enabled by default. To disable it, set auto_schema = 0
explicitly in your configuration. To re-enable it, set auto_schema = 1
or remove the auto_schema
setting from the configuration.
Keep in mind that the /bulk
HTTP endpoint does not support automatic table creation.
- Disable
- Enable
auto_schema = 0 # disable automatic table creation
This setting controls the binary log transaction flush/sync mode. It is optional, with a default value of 2 (flush every transaction, sync every second).
The directive determines how frequently the binary log will be flushed to the OS and synced to disk. There are three supported modes:
- 0, flush and sync every second. This offers the best performance, but up to 1 second worth of committed transactions can be lost in the event of a server crash or an OS/hardware crash.
- 1, flush and sync every transaction. This mode provides the worst performance but guarantees that every committed transaction's data is saved.
- 2, flush every transaction, sync every second. This mode delivers good performance and ensures that every committed transaction is saved in case of a server crash. However, in the event of an OS/hardware crash, up to 1 second worth of committed transactions can be lost.
For those familiar with MySQL and InnoDB, this directive is similar to innodb_flush_log_at_trx_commit
. In most cases, the default hybrid mode 2 provides a nice balance of speed and safety, with full RT table data protection against server crashes and some protection against hardware ones.
- Example
binlog_flush = 1 # ultimate safety, low speed
This setting controls the maximum binary log file size. It is optional, with a default value of 268435456, or 256 MB.
A new binlog file will be forcibly opened once the current binlog file reaches this size limit. This results in a finer granularity of logs and can lead to more efficient binlog disk usage under certain borderline workloads. A value of 0 indicates that the binlog file should not be reopened based on size.
- Example
binlog_max_log_size = 16M
This setting determines the path for binary log (also known as transaction log) files. It is optional, with a default value of the build-time configured data directory (e.g., /var/lib/manticore/data/binlog.*
in Linux).
Binary logs are used for crash recovery of RT table data and for attribute updates of plain disk indices that would otherwise only be stored in RAM until flush. When logging is enabled, every transaction COMMIT-ted into an RT table is written into a log file. Logs are then automatically replayed on startup after an unclean shutdown, recovering the logged changes.
The binlog_path
directive specifies the location of binary log files. It should only contain the path; searchd
will create and unlink multiple binlog.*
files in the directory as necessary (including binlog data, metadata, and lock files, etc).
An empty value disables binary logging, which improves performance but puts the RT table data at risk.
- Example
binlog_path = # disable logging
binlog_path = /var/lib/manticore/data # /var/lib/manticore/data/binlog.001 etc will be created
This setting determines the path to the Manticore Buddy binary. It is optional, with a default value being the build-time configured path, which varies across different operating systems. Typically, you don't need to modify this setting. However, it may be useful if you wish to run Manticore Buddy in debug mode, make changes to Manticore Buddy, or implement a new plugin. In the latter case, you can git clone
Buddy from https://github.com/manticoresoftware/manticoresearch-buddy, add a new plugin to the directory ./plugins/
, and run composer install --prefer-source
for easier development after you change the directory to the Buddy source.
To ensure you can run composer
, your machine must have PHP 8.2 or higher installed with the following extensions:
--enable-dom
--with-libxml
--enable-tokenizer
--enable-xml
--enable-xmlwriter
--enable-xmlreader
--enable-simplexml
--enable-phar
--enable-bcmath
--with-gmp
--enable-debug
--with-mysqli
--enable-mysqlnd
You can also opt for the special manticore-executor-dev
version for Linux amd64 available in the releases, for example: https://github.com/manticoresoftware/executor/releases/tag/v1.0.13
If you go this route, remember to link the dev version of the manticore executor to /usr/bin/php
.
- Example
buddy_path = manticore-executor -n /usr/share/manticore/modules/manticore-buddy/src/main.php --debug # use the default Manticore Buddy in Linux, but run it in debug mode
buddy_path = manticore-executor -n /opt/homebrew/share/manticore/modules/manticore-buddy/bin/manticore-buddy/src/main.php --debug # use the default Manticore Buddy in MacOS arm64, but run it in debug mode
buddy_path = manticore-executor -n /Users/username/manticoresearch-buddy/src/main.php --debug # use Manticore Buddy from a non-default location
This setting determines the maximum time to wait between requests (in seconds or special_suffixes) when using persistent connections. It is optional, with a default value of five minutes.
- Example
client_timeout = 1h
Server libc locale. Optional, default is C.
Specifies the libc locale, affecting the libc-based collations. Refer to collations section for the details.
- Example
collation_libc_locale = fr_FR
Default server collation. Optional, default is libc_ci.
Specifies the default collation used for incoming requests. The collation can be overridden on a per-query basis. Refer to collations section for the list of available collations and other details.
- Example
collation_server = utf8_ci
When specified, this setting enables the real-time mode, which is an imperative way of managing data schema. The value should be a path to the directory where you want to store all your tables, binary logs, and everything else needed for the proper functioning of Manticore Search in this mode.
Indexing of plain tables is not allowed when the data_dir
is specified. Read more about the difference between the RT mode and the plain mode in this section.
- Example
data_dir = /var/lib/manticore
This setting specifies the maximum size of document blocks from document storage that are held in memory. It is optional, with a default value of 16m (16 megabytes).
When stored_fields
is used, document blocks are read from disk and uncompressed. Since every block typically holds several documents, it may be reused when processing the next document. For this purpose, the block is held in a server-wide cache. The cache holds uncompressed blocks.
- Example
docstore_cache_size = 8m
Default attribute storage engine used when creating tables in RT mode. Can be rowwise
(default) or columnar
.
- Example
engine = columnar
This setting determines the maximum number of expanded keywords for a single wildcard. It is optional, with a default value of 0 (no limit).
When performing substring searches against tables built with dict = keywords
enabled, a single wildcard may potentially result in thousands or even millions of matched keywords (think of matching a*
against the entire Oxford dictionary). This directive allows you to limit the impact of such expansions. Setting expansion_limit = N
restricts expansions to no more than N of the most frequent matching keywords (per each wildcard in the query).
- Example
expansion_limit = 16
This setting determines the maximum number of documents in the expanded keyword that allows merging all such keywords together. It is optional, with a default value of 32.
When performing substring searches against tables built with dict = keywords
enabled, a single wildcard may potentially result in thousands or even millions of matched keywords. This directive allows you to increase the limit of how many keywords will merge together to speed up matching but uses more memory in the search.
- Example
expansion_merge_threshold_docs = 1024
This setting determines the maximum number of hits in the expanded keyword that allows merging all such keywords together. It is optional, with a default value of 256.
When performing substring searches against tables built with dict = keywords
enabled, a single wildcard may potentially result in thousands or even millions of matched keywords. This directive allows you to increase the limit of how many keywords will merge together to speed up matching but uses more memory in the search.
- Example
expansion_merge_threshold_hits = 512
This setting specifies whether timed grouping in API and SQL will be calculated in the local timezone or in UTC. It is optional, with a default value of 0 (meaning 'local timezone').
By default, all 'group by time' expressions (like group by day, week, month, and year in API, also group by day, month, year, yearmonth, yearmonthday in SQL) are done using local time. For example, if you have documents with attributes timed 13:00 utc
and 15:00 utc
, in the case of grouping, they both will fall into facility groups according to your local timezone setting. If you live in utc
, it will be one day, but if you live in utc+10
, then these documents will be matched into different group by day
facility groups (since 13:00 utc in UTC+10 timezone is 23:00 local time, but 15:00 is 01:00 of the next day). Sometimes such behavior is unacceptable, and it is desirable to make time grouping not dependent on timezone. You can run the server with a defined global TZ environment variable, but it will affect not only grouping but also timestamping in the logs, which may be undesirable as well. Switching 'on' this option (either in config or using SET global statement in SQL) will cause all time grouping expressions to be calculated in UTC, leaving the rest of time-depentend functions (i.e. logging of the server) in local TZ.
This setting specifies the timezone to be used by date/time-related functions. By default, the local timezone is used, but you can specify a different timezone in IANA format (e.g., Europe/Amsterdam
).
Note that this setting has no impact on logging, which always operates in the local timezone.
Also, note that if grouping_in_utc
is used, the 'group by time' function will still use UTC, while other date/time-related functions will use the specified timezone. Overall, it is not recommended to mix grouping_in_utc
and timezone
.
You can configure this option either in the config or by using the SET global statement in SQL.
This setting specifies the agent mirror statistics window size, in seconds (or special_suffixes). It is optional, with a default value of 60 seconds.
For a distributed table with agent mirrors in it (see more in agent, the master tracks several different per-mirror counters. These counters are then used for failover and balancing (the master picks the best mirror to use based on the counters). Counters are accumulated in blocks of ha_period_karma
seconds.
After beginning a new block, the master may still use the accumulated values from the previous one until the new one is half full. As a result, any previous history stops affecting the mirror choice after 1.5 times ha_period_karma seconds at most.
Even though at most two blocks are used for mirror selection, up to 15 last blocks are stored for instrumentation purposes. These blocks can be inspected using the SHOW AGENT STATUS statement.
- Example
ha_period_karma = 2m
This setting configures the interval between agent mirror pings, in milliseconds (or special_suffixes). It is optional, with a default value of 1000 milliseconds.
For a distributed table with agent mirrors in it (see more in agent), the master sends all mirrors a ping command during idle periods. This is to track the current agent status (alive or dead, network roundtrip, etc). The interval between such pings is defined by this directive. To disable pings, set ha_ping_interval to 0.
- Example
ha_ping_interval = 3s
The hostname_lookup
option defines the strategy for renewing hostnames. By default, the IP addresses of agent host names are cached at server start to avoid excessive access to DNS. However, in some cases, the IP can change dynamically (e.g. cloud hosting) and it may be desirable to not cache the IPs. Setting this option to request
disables the caching and queries the DNS for each query. The IP addresses can also be manually renewed using the FLUSH HOSTNAMES
command.
The jobs_queue_size setting defines how many "jobs" can be in the queue at the same time. It is unlimited by default.
In most cases, a "job" means one query to a single local table (plain table or a disk chunk of a real-time table). For example, if you have a distributed table consisting of 2 local tables or a real-time table with 2 disk chunks, a search query to either of them will mostly put 2 jobs in the queue. Then, the thread pool (whose size is defined by threads will process them. However, in some cases, if the query is too complex, more jobs can be created. Changing this setting is recommended when max_connections and threads are not enough to find a balance between the desired performance.
The listen_backlog setting determines the length of the TCP listen backlog for incoming connections. This is particularly relevant for Windows builds that process requests one by one. When the connection queue reaches its limit, new incoming connections will be refused. For non-Windows builds, the default value should work fine, and there is usually no need to adjust this setting.
- Example
listen_backlog = 20
This setting lets you specify an IP address and port, or Unix-domain socket path, that Manticore will accept connections on.
The general syntax for listen
is:
listen = ( address ":" port | port | path | address ":" port start - port end ) [ ":" protocol [ "_vip" ] [ "_readonly" ] ]
You can specify:
- either an IP address (or hostname) and a port number
- or just a port number
- or a Unix socket path (not supported on Windows)
- or an IP address and port range
If you specify a port number but not an address, searchd
will listen on all network interfaces. Unix path is identified by a leading slash. Port range can be set only for the replication protocol.
You can also specify a protocol handler (listener) to be used for connections on this socket. The listeners are:
- Not specified - Manticore will accept connections at this port from:
- other Manticore agents (i.e., a remote distributed table)
- clients via HTTP and HTTPS
- Manticore Buddy. Ensure you have a listener of this kind (or an
http
listener, as mentioned below) to avoid limitations in Manticore functionality.
mysql
MySQL protocol for connections from MySQL clients. Note:- Compressed protocol is also supported.
- If SSL is enabled, you can make an encrypted connection.
replication
- replication protocol used for nodes communication. More details can be found in the replication section. You can specify multiple replication listeners, but they must all listen on the same IP; only the ports can be different.http
- same as Not specified. Manticore will accept connections at this port from remote agents and clients via HTTP and HTTPS.https
- HTTPS protocol. Manticore will accept only HTTPS connections at this port. More details can be found in section SSL.sphinx
- legacy binary protocol. Used to serve connections from remote SphinxSE clients. Some Sphinx API clients implementations (an example is the Java one) require the explicit declaration of the listener.
Adding suffix _vip
to client protocols (that is, all except replication
, for instance mysql_vip
or http_vip
or just _vip
) forces creating a dedicated thread for the connection to bypass different limitations. That's useful for node maintenance in case of severe overload when the server would either stall or not let you connect via a regular port otherwise.
Suffix _readonly
sets read-only mode for the listener and limits it to accept only read queries.
- Example
listen = localhost
listen = localhost:5000 # listen for remote agents (binary API) and http/https requests on port 5000 at localhost
listen = 192.168.0.1:5000 # listen for remote agents (binary API) and http/https requests on port 5000 at 192.168.0.1
listen = /var/run/manticore/manticore.s # listen for binary API requests on unix socket
listen = /var/run/manticore/manticore.s:mysql # listen for mysql requests on unix socket
listen = 9312 # listen for remote agents (binary API) and http/https requests on port 9312 on any interface
listen = localhost:9306:mysql # listen for mysql requests on port 9306 at localhost
listen = localhost:9307:mysql_readonly # listen for mysql requests on port 9307 at localhost and accept only read queries
listen = 127.0.0.1:9308:http # listen for http requests as well as connections from remote agents (and binary API) on port 9308 at localhost
listen = 192.168.0.1:9320-9328:replication # listen for replication connections on ports 9320-9328 at 192.168.0.1
listen = 127.0.0.1:9443:https # listen for https requests (not http) on port 9443 at 127.0.0.1
listen = 127.0.0.1:9312:sphinx # listen for legacy Sphinx requests (e.g. from SphinxSE) on port 9312 at 127.0.0.1
There can be multiple listen
directives. searchd
will listen for client connections on all specified ports and sockets. The default config provided in Manticore packages defines listening on ports:
9308
and9312
for connections from remote agents and non-MySQL based clients- and on port
9306
for MySQL connections.
If you don't specify any listen
in the configuration at all, Manticore will wait for connections on:
127.0.0.1:9306
for MySQL clients127.0.0.1:9312
for HTTP/HTTPS and connections from other Manticore nodes and clients based on the Manticore binary API.
By default, Linux won't allow you to let Manticore listen on a port below 1024 (e.g. listen = 127.0.0.1:80:http
or listen = 127.0.0.1:443:https
) unless you run searchd under root. If you still want to be able to start Manticore, so it listens on ports < 1024 under a non-root user, consider doing one of the following (either of these should work):
- Run the command
setcap CAP_NET_BIND_SERVICE=+eip /usr/bin/searchd
- Add
AmbientCapabilities=CAP_NET_BIND_SERVICE
to Manticore's systemd unit and reload the daemon (systemctl daemon-reload
).
This setting allows the TCP_FASTOPEN flag for all listeners. By default, it is managed by the system but may be explicitly switched off by setting it to '0'.
For general knowledge about the TCP Fast Open extension, please consult with Wikipedia. In short, it allows the elimination of one TCP round-trip when establishing a connection.
In practice, using TFO in many situations may optimize client-agent network efficiency, as if persistent agents are in play, but without holding active connections, and also without limitation for the maximum num of connections.
On modern OS, TFO support is usually switched 'on' at the system level, but this is just a 'capability', not the rule. Linux (as the most progressive) has supported it since 2011, on kernels starting from 3.7 (for the server-side). Windows has supported it from some builds of Windows 10. Other operating systems (FreeBSD, MacOS) are also in the game.
For Linux system server checks variable /proc/sys/net/ipv4/tcp_fastopen
and behaves according to it. Bit 0 manages client side, bit 1 rules listeners. By default, the system has this parameter set to 1, i.e., clients enabled, listeners disabled.
The log setting specifies the name of the log file where all searchd
run time events will be logged. If not specified, the default name is 'searchd.log'.
Alternatively, you can use the 'syslog' as the file name. In this case, the events will be sent to the syslog daemon. To use the syslog option, you need to configure Manticore with the -–with-syslog
option during building.
- Example
log = /var/log/searchd.log
Limits the amount of queries per batch. Optional, default is 32.
Makes searchd perform a sanity check of the amount of queries submitted in a single batch when using multi-queries. Set it to 0 to skip the check.
- Example
max_batch_queries = 256
Maximum number of simultaneous client connections. Unlimited by default. That is usually noticeable only when using any kind of persistent connections, like cli mysql sessions or persistent remote connections from remote distributed tables. When the limit is exceeded you can still connect to the server using the VIP connection. VIP connections are not counted towards the limit.
- Example
max_connections = 10
Instance-wide limit of threads one operation can use. By default, appropriate operations can occupy all CPU cores, leaving no room for other operations. For example, call pq
against a considerably large percolate table can utilize all threads for tens of seconds. Setting max_threads_per_query
to, say, half of threads will ensure that you can run a couple of such call pq
operations in parallel.
You can also set this setting as a session or a global variable during runtime.
Additionally, you can control the behavior on a per-query basis with the help of the threads OPTION.
- Example
max_threads_per_query = 4
Maximum allowed per-query filter count. This setting is only used for internal sanity checks and does not directly affect RAM usage or performance. Optional, the default is 256.
- Example
max_filters = 1024
Maximum allowed per-filter values count. This setting is only used for internal sanity checks and does not directly affect RAM usage or performance. Optional, the default is 4096.
- Example
max_filter_values = 16384
The maximum number of files that the server is allowed to open is called the "soft limit". Note that serving large fragmented real-time tables may require this limit to be set high, as each disk chunk may occupy a dozen or more files. For example, a real-time table with 1000 chunks may require thousands of files to be opened simultaneously. If you encounter the error 'Too many open files' in the logs, try adjusting this option, as it may help resolve the issue.
There is also a "hard limit" that cannot be exceeded by the option. This limit is defined by the system and can be changed in the file /etc/security/limits.conf
on Linux. Other operating systems may have different approaches, so consult your manuals for more information.
- Example
max_open_files = 10000
Apart from direct numeric values, you can use the magic word 'max' to set the limit equal to the available current hard limit.
- Example
max_open_files = max
Maximum allowed network packet size. This setting limits both query packets from clients and response packets from remote agents in a distributed environment. Only used for internal sanity checks, it does not directly affect RAM usage or performance. Optional, the default is 8M.
- Example
max_packet_size = 32M
A server version string to return via the MySQL protocol. Optional, the default is empty (returns the Manticore version).
Several picky MySQL client libraries depend on a particular version number format used by MySQL, and moreover, sometimes choose a different execution path based on the reported version number (rather than the indicated capabilities flags). For instance, Python MySQLdb 1.2.2 throws an exception when the version number is not in X.Y.ZZ format; MySQL .NET connector 6.3.x fails internally on version numbers 1.x along with a certain combination of flags, etc. To work around that, you can use the mysql_version_string
directive and have searchd
report a different version to clients connecting over the MySQL protocol. (By default, it reports its own version.)
- Example
mysql_version_string = 5.0.37
Number of network threads, the default is 1.
This setting is useful for extremely high query rates when just one thread is not enough to manage all the incoming queries.
Controls the busy loop interval of the network thread. The default is -1, and it can be set to -1, 0, or a positive integer.
In cases where the server is configured as a pure master and just routes requests to agents, it is important to handle requests without delays and not allow the network thread to sleep. There is a busy loop for that. After an incoming request, the network thread uses CPU poll for 10 * net_wait_tm
milliseconds if net_wait_tm
is a positive number or polls only with the CPU if net_wait_tm
is 0
. Also, the busy loop can be disabled with net_wait_tm = -1
- in this case, the poller sets the timeout to the actual agent's timeouts on the system polling call.
WARNING: A CPU busy loop actually loads the CPU core, so setting this value to any non-default value will cause noticeable CPU usage even with an idle server.
Defines how many clients are accepted on each iteration of the network loop. Default is 0 (unlimited), which should be fine for most users. This is a fine-tuning option to control the throughput of the network loop in high load scenarios.
Defines how many requests are processed on each iteration of the network loop. The default is 0 (unlimited), which should be fine for most users. This is a fine-tuning option to control the throughput of the network loop in high load scenarios.
Network client request read/write timeout, in seconds (or special_suffixes). Optional, the default is 5 seconds. searchd
will forcibly close a client connection which fails to send a query or read a result within this timeout.
Note also the reset_network_timeout_on_packet parameter. This parameter alters the behavior of network_timeout
from applying to the entire query
or result
to individual packets instead. Typically, a query/result fits within one or two packets. However, in cases where a large amount of data is required, this parameter can be invaluable in maintaining active operations.
- Example
network_timeout = 10s
This setting allows you to specify the network address of the node. By default, it is set to the replication listen ddress. This is correct in most cases; however, there are situations where you have to specify it manually:
- Node behind a firewall
- Network address translation enabled (NAT)
- Container deployments, such as Docker or cloud deployments
- Clusters with nodes in more than one region
- Example
node_address = 10.101.0.10
This setting determines whether to allow queries with only the negation full-text operator. Optional, the default is 0 (fail queries with only the NOT operator).
- Example
not_terms_only_allowed = 1
Sets the default table compaction threshold. Read more here - Number of optimized disk chunks. This setting can be overridden with the per-query option cutoff. It can also be changed dynamically via SET GLOBAL.
- Example
optimize_cutoff = 4
This setting determines the maximum number of simultaneous persistent connections to remote persistent agents. Each time an agent defined under agent_persistent
is connected, we try to reuse an existing connection (if any), or connect and save the connection for future use. However, in some cases, it makes sense to limit the number of such persistent connections. This directive defines the limit. It affects the number of connections to each agent's host across all distributed tables.
It is reasonable to set the value equal to or less than the max_connections option in the agent's config.
- Example
persistent_connections_limit = 29 # assume that each host of agents has max_connections = 30 (or 29).
pid_file is a mandatory configuration option in Manticore search that specifies the path of the file where the process ID of the searchd
server is stored.
The searchd process ID file is re-created and locked on startup, and contains the head server process ID while the server is running. It is unlinked on server shutdown.
The purpose of this file is to enable Manticore to perform various internal tasks, such as checking whether there is already a running instance of searchd
, stopping searchd
, and notifying it that it should rotate the tables. The file can also be used for external automation scripts.
- Example
pid_file = /var/run/manticore/searchd.pid
Costs for the query time prediction model, in nanoseconds. Optional, the default is doc=64, hit=48, skip=2048, match=64
.
- Example
predicted_time_costs = doc=128, hit=96, skip=4096, match=128
Terminating queries before completion based on their execution time (with the max query time setting) is a nice safety net, but it comes with an inherent drawback: indeterministic (unstable) results. That is, if you repeat the very same (complex) search query with a time limit several times, the time limit will be hit at different stages, and you will get different result sets.
- SQL
- API
SELECT … OPTION max_query_time
There is a new option, SELECT … OPTION max_predicted_time, that lets you limit the query time and get stable, repeatable results. Instead of regularly checking the actual current time while evaluating the query, which is indeterministic, it predicts the current running time using a simple linear model instead:
predicted_time =
doc_cost * processed_documents +
hit_cost * processed_hits +
skip_cost * skiplist_jumps +
match_cost * found_matches
The query is then terminated early when the predicted_time
reaches a given limit.
Of course, this is not a hard limit on the actual time spent (it is, however, a hard limit on the amount of processing work done), and a simple linear model is in no way an ideally precise one. So the wall clock time may be either below or over the target limit. However, the error margins are quite acceptable: for instance, in our experiments with a 100 msec target limit, the majority of the test queries fell into a 95 to 105 msec range, and all the queries were in an 80 to 120 msec range. Also, as a nice side effect, using the modeled query time instead of measuring the actual run time results in somewhat fewer gettimeofday() calls, too.
No two server makes and models are identical, so the predicted_time_costs
directive lets you configure the costs for the model above. For convenience, they are integers, counted in nanoseconds. (The limit in max_predicted_time is counted in milliseconds, and having to specify cost values as 0.000128 ms instead of 128 ns is somewhat more error-prone.) It is not necessary to specify all four costs at once, as the missed ones will take the default values. However, we strongly suggest specifying all of them for readability.
The preopen_tables configuration directive specifies whether to forcibly preopen all tables on startup. The default value is 1, which means that all tables will be preopened regardless of the per-table preopen setting. If set to 0, the per-table settings can take effect, and they will default to 0.
Pre-opening tables can prevent races between search queries and rotations that can cause queries to fail occasionally. However, it also uses more file handles. In most scenarios, it is recommended to preopen tables.
Here's an example configuration:
- Example
preopen_tables = 1
The pseudo_sharding configuration option enables parallelization of search queries to local plain and real-time tables, regardless of whether they are queried directly or through a distributed table. This feature will automatically parallelize queries to up to the number of threads specified in searchd.threads
# of threads.
Note that if your worker threads are already busy, because you have:
- high query concurrency
- physical sharding of any kind:
- distributed table of multiple plain/real-time tables
- real-time table consisting of too many disk chunks
then enabling pseudo_sharding may not provide any benefits and may even result in a slight decrease in throughput. If you prioritize higher throughput over lower latency, it's recommended to disable this option.
Enabled by default.
- Example
pseudo_sharding = 0
The replication_connect_timeout
directive defines the timeout for connecting to a remote node. By default, the value is assumed to be in milliseconds, but it can have another suffix. The default value is 1000 (1 second).
When connecting to a remote node, Manticore will wait for this amount of time at most to complete the connection successfully. If the timeout is reached but the connection has not been established, and retries
are enabled, a retry will be initiated.
The replication_query_timeout
sets the amount of time that searchd will wait for a remote node to complete a query. The default value is 3000 milliseconds (3 seconds), but can be suffixed
to indicate a different unit of time.
After establishing a connection, Manticore will wait for a maximum of replication_query_timeout
for the remote node to complete. Note that this timeout is separate from the replication_connect_timeout
, and the total possible delay caused by a remote node will be the sum of both values.
This setting is an integer that specifies how many times Manticore will attempt to connect and query a remote node during replication before reporting a fatal query error. The default value is 3.
This setting is an integer in milliseconds (or special_suffixes) that specifies the delay before Manticore retries querying a remote node in case of failure during replication. This value is only relevant when a non-zero value is specified. The default value is 500.
This configuration sets the maximum amount of RAM allocated for cached result sets in bytes. The default value is 16777216, which is equivalent to 16 megabytes. If the value is set to 0, the query cache is disabled. For more information about the query cache, please refer to the query cache for details.
- Example
qcache_max_bytes = 16777216
Integer, in milliseconds. The minimum wall time threshold for a query result to be cached. Defaults to 3000, or 3 seconds. 0 means cache everything. Refer to query cache for details. This value also may be expressed with time special_suffixes, but use it with care and don't confuse yourself with the name of the value itself, containing '_msec'.
Integer, in seconds. The expiration period for a cached result set. Defaults to 60, or 1 minute. The minimum possible value is 1 second. Refer to query cache for details. This value also may be expressed with time special_suffixes, but use it with care and don't confuse yourself with the name of the value itself, containing '_sec'.
Query log format. Optional, allowed values are plain
and sphinxql
, default is sphinxql
.
The sphinxql
mode logs valid SQL statements. The plain
mode logs queries in a plain text format (mostly suitable for purely full-text use cases). This directive allows you to switch between the two formats on search server startup. The log format can also be altered on the fly, using SET GLOBAL query_log_format=sphinxql
syntax. Refer to Query logging for more details.
- Example
query_log_format = sphinxql
Limit (in milliseconds) that prevents the query from being written to the query log. Optional, default is 0 (all queries are written to the query log). This directive specifies that only queries with execution times that exceed the specified limit will be logged (this value also may be expressed with time special_suffixes, but use it with care and don't confuse yourself with the name of the value itself, containing _msec
).
Query log file name. Optional, default is empty (do not log queries). All search queries (such as SELECT ... but not INSERT/REPLACE/UPDATE queries) will be logged in this file. The format is described in Query logging. In case of 'plain' format, you can use 'syslog' as the path to the log file. In this case, all search queries will be sent to the syslog daemon with LOG_INFO
priority, prefixed with '[query]' instead of timestamp. To use the syslog option, Manticore must be configured with -–with-syslog
on building.
- Example
query_log = /var/log/query.log
The query_log_mode directive allows you to set a different permission for the searchd and query log files. By default, these log files are created with 600 permission, meaning that only the user under which the server runs and root users can read the log files. This directive can be handy if you want to allow other users to read the log files, for example, monitoring solutions running on non-root users.
- Example
query_log_mode = 666
The read_buffer_docs directive controls the per-keyword read buffer size for document lists. For every keyword occurrence in every search query, there are two associated read buffers: one for the document list and one for the hit list. This setting lets you control the document list buffer size.
A larger buffer size might increase per-query RAM use, but it could possibly decrease I/O time. It makes sense to set larger values for slow storage, but for storage capable of high IOPS, experimenting should be done in the low values area.
The default value is 256K, and the minimal value is 8K. You may also set read_buffer_docs on a per-table basis, which will override anything set on the server's config level.
- Example
read_buffer_docs = 128K
The read_buffer_hits directive specifies the per-keyword read buffer size for hit lists in search queries. By default, the size is 256K and the minimum value is 8K. For every keyword occurrence in a search query, there are two associated read buffers, one for the document list and one for the hit list. Increasing the buffer size can increase per-query RAM use but decrease I/O time. For slow storage, larger buffer sizes make sense, while for storage capable of high IOPS, experimenting should be done in the low values area.
This setting can also be specified on a per-table basis using the read_buffer_hits option in read_buffer_hits which will override the server-level setting.
- Example
read_buffer_hits = 128K
Unhinted read size. Optional, default is 32K, minimal 1K
When querying, some reads know in advance exactly how much data is there to be read, but some currently do not. Most prominently, hit list size is not currently known in advance. This setting lets you control how much data to read in such cases. It impacts hit list I/O time, reducing it for lists larger than unhinted read size, but raising it for smaller lists. It does not affect RAM usage because the read buffer will already be allocated. So it should not be greater than read_buffer.
- Example
read_unhinted = 32K
Refines the behavior of networking timeouts (such as network_timeout
, read_timeout
, and agent_query_timeout
).
When set to 0, timeouts limit the maximum time for sending the entire request/query. When set to 1 (default), timeouts limit the maximum time between network activities.
With replication, a node may need to send a large file (for example, 100GB) to another node. Assume the network can transfer data at 1GB/s, with a series of packets of 4-5MB each. To transfer the entire file, you would need 100 seconds. A default timeout of 5 seconds would only allow the transfer of 5GB before the connection is dropped. Increasing the timeout could be a workaround, but it is not scalable (for instance, the next file might be 150GB, leading to failure again). However, with the default reset_network_timeout_on_packet
set to 1, the timeout is applied not to the entire transfer but to individual packets. As long as the transfer is in progress (and data is actually being received over the network during the timeout period), it is kept alive. If the transfer gets stuck, such that a timeout occurs between packets, it will be dropped.
Note that if you set up a distributed table, each node — both master and agents — should be tuned. On the master side, agent_query_timeout
is affected; on agents, network_timeout
is relevant.
- Example
reset_network_timeout_on_packet = 0
RT tables RAM chunk flush check period, in seconds (or special_suffixes). Optional, default is 10 hours.
Actively updated RT tables that fully fit in RAM chunks can still result in ever-growing binlogs, impacting disk use and crash recovery time. With this directive, the search server performs periodic flush checks, and eligible RAM chunks can be saved, enabling consequential binlog cleanup. See Binary logging for more details.
- Example
rt_flush_period = 3600 # 1 hour
A maximum number of I/O operations (per second) that the RT chunks merge thread is allowed to start. Optional, default is 0 (no limit).
This directive lets you throttle down the I/O impact arising from the OPTIMIZE
statements. It is guaranteed that all RT optimization activities will not generate more disk IOPS (I/Os per second) than the configured limit. Limiting rt_merge_iops can reduce search performance degradation caused by merging.
- Example
rt_merge_iops = 40
A maximum size of an I/O operation that the RT chunks merge thread is allowed to start. Optional, default is 0 (no limit).
This directive lets you throttle down the I/O impact arising from the OPTIMIZE
statements. I/Os larger than this limit will be broken down into two or more I/Os, which will then be accounted for as separate I/Os with regards to the rt_merge_iops limit. Thus, it is guaranteed that all optimization activities will not generate more than (rt_merge_iops * rt_merge_maxiosize) bytes of disk I/O per second.
- Example
rt_merge_maxiosize = 1M
Prevents searchd
stalls while rotating tables with huge amounts of data to precache. Optional, default is 1 (enable seamless rotation). On Windows systems, seamless rotation is disabled by default.
Tables may contain some data that needs to be precached in RAM. At the moment, .spa
, .spb
, .spi
, and .spm
files are fully precached (they contain attribute data, blob attribute data, keyword table, and killed row map, respectively.) Without seamless rotate, rotating a table tries to use as little RAM as possible and works as follows:
- New queries are temporarily rejected (with "retry" error code);
searchd
waits for all currently running queries to finish;- The old table is deallocated, and its files are renamed;
- New table files are renamed, and required RAM is allocated;
- New table attribute and dictionary data are preloaded to RAM;
searchd
resumes serving queries from the new table.
However, if there's a lot of attribute or dictionary data, then the preloading step could take a noticeable amount of time - up to several minutes in the case of preloading 1-5+ GB files.
With seamless rotate enabled, rotation works as follows:
- New table RAM storage is allocated;
- New table attribute and dictionary data are asynchronously preloaded to RAM;
- On success, the old table is deallocated, and both tables' files are renamed;
- On failure, the new table is deallocated;
- At any given moment, queries are served either from the old or new table copy.
Seamless rotate comes at the cost of higher peak memory usage during the rotation (because both old and new copies of .spa/.spb/.spi/.spm
data need to be in RAM while preloading the new copy). Average usage remains the same.
- Example
seamless_rotate = 1
This option enables/disables the use of secondary indexes for search queries. It is optional, and the default is 1 (enabled). Note that you don't need to enable it for indexing as it is always enabled as long as the Manticore Columnar Library is installed. The latter is also required for using the indexes when searching. There are three modes available:
0
: Disable the use of secondary indexes on search. They can be enabled for individual queries using analyzer hints1
: Enable the use of secondary indexes on search. They can be disabled for individual queries using analyzer hintsforce
: Same as enable, but any errors during the loading of secondary indexes will be reported, and the whole index will not be loaded into the daemon.
Note that secondary indexes are not effective for full-text queries.
- Example
secondary_indexes = 1
Integer number that serves as a server identifier used as a seed to generate a unique short UUID for nodes that are part of a replication cluster. The server_id must be unique across the nodes of a cluster and in the range from 0 to 127. If server_id is not set, the MAC address or a random number will be used as a seed for the short UUID.
- Example
server_id = 1
searchd --stopwait
waiting time, in seconds (or special_suffixes). Optional, default is 60 seconds.
When you run searchd --stopwait
your server needs to perform some activities before stopping, such as finishing queries, flushing RT RAM chunks, flushing attributes, and updating the binlog. These tasks require some time. searchd --stopwait
will wait up to shutdown_time
seconds for the server to finish its jobs. The suitable time depends on your table size and load.
- Example
shutdown_timeout = 3m # wait for up to 3 minutes
SHA1 hash of the password required to invoke the 'shutdown' command from a VIP Manticore SQL connection. Without it,debug 'shutdown' subcommand will never cause the server to stop. Note that such simple hashing should not be considered strong protection, as we don't use a salted hash or any kind of modern hash function. It is intended as a fool-proof measure for housekeeping daemons in a local network.
A prefix to prepend to the local file names when generating snippets. Optional, default is the current working folder.
This prefix can be used in distributed snippets generation along with load_files
or load_files_scattered
options.
Note that this is a prefix and not a path! This means that if a prefix is set to "server1" and the request refers to "file23", searchd
will attempt to open "server1file23" (all of that without quotes). So, if you need it to be a path, you have to include the trailing slash.
After constructing the final file path, the server unwinds all relative dirs and compares the final result with the value of snippet_file_prefix
. If the result does not begin with the prefix, such a file will be rejected with an error message.
For example, if you set it to /mnt/data
and someone calls snippet generation with the file ../../../etc/passwd
as the source, they will get the error message:
File '/mnt/data/../../../etc/passwd' escapes '/mnt/data/' scope
instead of the content of the file.
Also, with a non-set parameter and reading /etc/passwd
, it will actually read /daemon/working/folder/etc/passwd since the default for the parameter is the server's working folder.
Note also that this is a local option; it does not affect the agents in any way. So you can safely set a prefix on a master server. The requests routed to the agents will not be affected by the master's setting. They will, however, be affected by the agent's own settings.
This might be useful, for instance, when the document storage locations (whether local storage or NAS mountpoints) are inconsistent across the servers.
- Example
snippets_file_prefix = /mnt/common/server1/
WARNING: If you still want to access files from the FS root, you have to explicitly set
snippets_file_prefix
to empty value (bysnippets_file_prefix=
line), or to root (bysnippets_file_prefix=/
).
Path to a file where the current SQL state will be serialized.
On server startup, this file gets replayed. On eligible state changes (e.g., SET GLOBAL), this file gets rewritten automatically. This can prevent a hard-to-diagnose problem: If you load UDF functions but Manticore crashes, when it gets (automatically) restarted, your UDF and global variables will no longer be available. Using persistent state helps ensure a graceful recovery with no such surprises.
sphinxql_state
cannot be used to execute arbitrary commands, such as CREATE TABLE
.
- Example
sphinxql_state = uservars.sql
Maximum time to wait between requests (in seconds, or special_suffixes) when using the SQL interface. Optional, default is 15 minutes.
- Example
sphinxql_timeout = 15m
Path to the SSL Certificate Authority (CA) certificate file (also known as root certificate). Optional, default is empty. When not empty, the certificate in ssl_cert
should be signed by this root certificate.
The server uses the CA file to verify the signature on the certificate. The file must be in PEM format.
- Example
ssl_ca = keys/ca-cert.pem
Path to the server's SSL certificate. Optional, default is empty.
The server uses this certificate as a self-signed public key to encrypt HTTP traffic over SSL. The file must be in PEM format.
- Example
ssl_cert = keys/server-cert.pem
Path to the SSL certificate key. Optional, default is empty.
The server uses this private key to encrypt HTTP traffic over SSL. The file must be in PEM format.
- Example
ssl_key = keys/server-key.pem
Max common subtree document cache size, per-query. Optional, default is 0 (disabled).
This setting limits the RAM usage of a common subtree optimizer (see multi-queries). At most, this much RAM will be spent to cache document entries for each query. Setting the limit to 0 disables the optimizer.
- Example
subtree_docs_cache = 8M
Max common subtree hit cache size, per-query. Optional, default is 0 (disabled).
This setting limits the RAM usage of a common subtree optimizer (see multi-queries). At most, this much RAM will be spent to cache keyword occurrences (hits) for each query. Setting the limit to 0 disables the optimizer.
- Example
subtree_hits_cache = 16M
Number of working threads (or, size of thread pool) for the Manticore daemon. Manticore creates this number of OS threads on start, and they perform all jobs inside the daemon, such as executing queries, creating snippets, etc. Some operations may be split into sub-tasks and executed in parallel, for example:
- Search in a real-time table
- Search in a distributed table consisting of local tables
- Percolate query call
- and others
By default, it's set to the number of CPU cores on the server. Manticore creates the threads on start and keeps them until it's stopped. Each sub-task can use one of the threads when it needs it. When the sub-task finishes, it releases the thread so another sub-task can use it.
In the case of intensive I/O type of load, it might make sense to set the value higher than the number of CPU cores.
- Example
threads = 10
Maximum stack size for a job (coroutine, one search query may cause multiple jobs/coroutines). Optional, default is 128K.
Each job has its own stack of 128K. When you run a query, it's checked for how much stack it requires. If the default 128K is enough, it's just processed. If it needs more, another job with an increased stack is scheduled, which continues processing. The maximum size of such an advanced stack is limited by this setting.
Setting the value to a reasonably high rate will help with processing very deep queries without implying that overall RAM consumption will grow too high. For example, setting it to 1G does not imply that every new job will take 1G of RAM, but if we see that it requires, let's say, 100M stack, we just allocate 100M for the job. Other jobs at the same time will be running with their default 128K stack. The same way, we can run even more complex queries that need 500M. And only if we see internally that the job requires more than 1G of stack, we will fail and report about too low thread_stack.
However, in practice, even a query which needs 16M of stack is often too complex for parsing and consumes too much time and resources to be processed. So, the daemon will process it, but limiting such queries by the thread_stack
setting looks quite reasonable.
- Example
thread_stack = 8M
Determines whether to unlink .old
table copies on successful rotation. Optional, default is 1 (do unlink).
- Example
unlink_old = 0
Threaded server watchdog. Optional, default is 1 (watchdog enabled).
When a Manticore query crashes, it can take down the entire server. With the watchdog feature enabled, searchd
also maintains a separate lightweight process that monitors the main server process and automatically restarts it in case of abnormal termination. The watchdog is enabled by default.
- Example
watchdog = 0 # disable watchdog
The lemmatizer_base
is an optional configuration directive that specifies the base path for lemmatizer dictionaries. The default path is /usr/share/manticore
The lemmatizer implementation in Manticore Search (see Morphology to learn what lemmatizers are) is dictionary-driven and requires specific dictionary files for different languages. These files can be downloaded from the Manticore website (https://manticoresearch.com/install/#other-downloads).
Example:
lemmatizer_base = /usr/share/manticore/
The progressive_merge is a configuration directive that, when enabled, merges real-time table disk chunks from smaller to larger ones. This approach speeds up the merging process and reduces read/write amplification. By default, this setting is enabled. If disabled, the chunks are merged in the order they were created.
The json_autoconv_keynames is an optional configuration directive that determines if and how to auto-convert key names within JSON attributes. The known value is 'lowercase'. By default, this setting is unspecified (meaning no conversion occurs).
When set to lowercase, key names within JSON attributes will be automatically converted to lowercase during indexing. This conversion applies to JSON attributes from all data sources, including SQL and XMLpipe2.
Example:
json_autoconv_keynames = lowercase
The json_autoconv_numbers is an optional configuration directive that determines whether to automatically detect and convert JSON strings that represent numbers into numeric attributes. The default value is 0 (do not convert strings into numbers).
When this option is set to 1, values such as "1234" will be indexed as numbers instead of strings. If the option is set to 0, such values will be indexed as strings. This conversion applies to JSON attributes from all data sources, including SQL and XMLpipe2.
Example:
json_autoconv_numbers = 1
on_json_attr_error is an optional configuration directive that specifies the action to take if JSON format errors are found. The default value is ignore_attr
(ignore errors). This setting applies only to sql_attr_json
attributes.
By default, JSON format errors are ignored (ignore_attr
), and the indexer tool will show a warning. Setting this option to fail_index
will cause indexing to fail at the first JSON format error.
Example:
on_json_attr_error = ignore_attr
The plugin_dir is an optional configuration directive that specifies the trusted location for dynamic libraries (UDFs). The default path is /usr/local/lib/manticore/
.
This directive sets the trusted directory from which the UDF libraries can be loaded.
Example:
plugin_dir = /usr/local/lib/manticore/
Manticore Search supports the use of special suffixes to simplify numeric values with specific meanings. These suffixes are categorized into size suffixes and time suffixes. The common format for suffixes is an integer
followed by a literal
, such as 10k
or 100d
. Literals are case-insensitive, so 10W
and 10w
are considered the same.
-
Size suffixes: These suffixes can be used in settings that define the size of something, such as memory buffer, disk file size, or RAM limit. If no suffix is specified, the value is considered in bytes by default. The available size suffixes are:
k
for kilobytes (1k = 1024 bytes)m
for megabytes (1m = 1024k)g
for gigabytes (1g = 1024m)t
for terabytes (1t = 1024g)
-
Time suffixes: These suffixes can be used in settings that define time interval values, such as delays or timeouts. Unadorned values for these parameters usually have a documented scale, but instead of guessing, you can use an explicit suffix. The available time suffixes are:
us
for microsecondsms
for millisecondss
for secondsm
for minutesh
for hoursd
for daysw
for weeks