About MyRocks for MariaDB

MyRocks is an open source storage engine that was originally developed by Facebook.

MyRocks has been extended by the MariaDB engineering team to be a pluggable storage engine that you use in your MariaDB solutions. It works seamlessly with MariaDB features. This openness in the storage layer allows you to use the right storage engine to optimize your usage requirements, which provides optimum performance. Community contributions are one of MariaDB’s greatest advantages over other databases. Under the lead of our developer Sergey Petrunia, MyRocks in MariaDB is occasionally being merged with upstream MyRocks from Facebook. See more at: facebook-myrocks-mariadb#sthash.ZlEr7kNq.dpuf

MyRocks, typically, gives greater performance for web scale type applications. It can be an ideal storage engine solution when you have workloads that require greater compression and IO efficiency. It uses a Log Structured Merge (LSM) architecture, which has advantages over B-Tree algorithms, to provide efficient data ingestion, like read-free replication slaves, or fast bulk data loading. MyRocks distinguishing features include:

• compaction filter

• merge operator

• backup

• column families

• bulk loading

• persistent cache

For more MyRocks features see: Features-Not-in-LevelDB

Benefits

On production workloads, MyRocks was tested to prove that it provides:

Greater Space Efficiency

• 2x more compression MyRocks has 2x better compression compared to compressed InnoDB, 3-4x better compression compared to uncompressed InnoDB, meaning you use less space.

Greater Writing Efficiency

• 2x lower write rates to storage MyRocks has a 10x less write amplification compared to InnoDB, giving you better endurance of flash storage and improving overall throughput.

Faster Data Loading

• faster database loads MyRocks writes data directly onto the bottommost level, which avoids all compaction overheads when you enable faster data loading for a session.

Faster Replication

• No random reads for updating secondary keys, except for unique indexes. The Read-Free Replication option does away with random reads when updating primary keys, regardless of uniqueness, with a row-based binary logging format.

myrocks.iofacebook-myrocks-mariadb

Requirements and Limitations

• MyRocks is included from MariaDB 10.2.5.

• MyRocks is available in the MariaDB Server packages for Linux and Windows.

• Maria DB optimistic parallel replication may not be supported.

• MyRocks is not available for 32-bit platforms

• Galera Cluster is tightly integrated into InnoDB storage engine (it also supports Percona's XtraDB which is a modified version of InnoDB). Galera Cluster does not work with any other storage engines, including MyRocks (or TokuDB for example).

MyRocks builds are available on platforms that support a sufficiently modern compiler, for example:

• Ubuntu Trusty, Xenial, (amd64 and ppc64el)

• Ubuntu Yakkety (amd64)

• Debian Jessie, stable (amd64, ppc64el)

• Debian Stretch, Sid (testing and unstable) (amd64)

• CentOS/RHEL 7 (amd64)

• Centos/RHEL 7.3 (amd64)

• Fedora 24 and 25 (amd64)

• OpenSUSE 42 (amd64)

• Windows 64 (zip and MSI)

_{This page is licensed: CC BY-SA / Gnu FDL}

This page describes how to get MyRocks in MariaDB when compiling MariaDB from source.

(See Build-Steps for instructions how to build the upstream)

Build Process and Requirements

MariaDB compile process will compile MyRocks into ha_rocksdb.so by default if the platform supports it (That is, no WITH_ROCKSDB switch is necessary).

Platform requirements:

• A 64-bit platform (due to some 32 bit compilers having difficulties with RocksDB)

• git installed (or git submodules fetched somehow)

• A sufficiently recent compiler:

  • gcc >= 4.8, or

  • clang >= 3.3, or

  • MS Visual Studio 2015 or newer

Building on Ubuntu 16.04

The steps were checked on a fresh install of Ubuntu 16.04.2 LTS Xenial.

sudo apt-get update
sudo apt-get -y install g++ cmake libbz2-dev libaio-dev bison zlib1g-dev libsnappy-dev 
sudo apt-get -y install libgflags-dev libreadline6-dev libncurses5-dev libssl-dev liblz4-dev gdb git
;

git clone https://github.com/MariaDB/server.git mariadb-10.2
cd mariadb-10.2
git checkout 10.2
git submodule init
git submodule update
cmake .
make -j10

This should produce storage/rocksdb/ha_rocksdb.so which is MyRocks storage engine in the loadable form.

Starting MyRocks

MyRocks does not require any special way to initialize the data directory. Minimal my.cnf flle:

cat > ~/my1.cnf <<EOF
[mysqld]

datadir=../mysql-test/var/install.db
plugin-dir=../storage/rocksdb
language=./share/english
socket=/tmp/mysql.sock
port=3307

plugin-load=ha_rocksdb
default-storage-engine=rocksdb
EOF

Run the server like this

(cd mysql-test; ./mtr alias)
cp -r mysql-test/var/install.db ~/data1
cd ../sql
./mysqld --defaults-file=~/my1.cnf

Compression libraries. Supported compression libraries are listed in rocksdb_supported_compression_types. Compiling like the above, I get:

Snappy,Zlib,LZ4,LZ4HC

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks is available in

• Facebook's (FB) MySQL branch (originally based on MySQL 5.6)

• MariaDB (from 10.2 and 10.3)

• Percona Server from 5.7

This page lists differences between these variants.

This is a work in progress. The contents are not final

RocksDB Data Location

FB and Percona store RocksdDB files in $datadir/.rocksdb. MariaDB puts them in $datadir/#rocksdb. This is more friendly for packaging and OS scripts.

Compression Algorithms

• FB's branch doesn't provide binaries. One needs to compile it with appropriate compression libraries.

• In MariaDB, available compression algorithms can be seen in the rocksdb_supported_compression_types variable. From MariaDB 10.7, algorithms can be installed as a plugin. In earlier versions, the set of supported compression algorithms depends on the platform.

  • On Ubuntu 16.04 (current LTS) it is Snappy,Zlib,LZ4,LZ4HC .

  • On CentOS 7.4 it is Snappy,Zlib.

  • In the bintar tarball it is Snappy,Zlib.

• Percona Server supports: Zlib, ZSTD, LZ4 (the default), LZ4HC. Unsupported algorithms: Snappy, BZip2, XPress.

RocksDB Version Information

• FB's branch provides the rocksdb_git_hash status variable.

• MariaDB provides the @@rocksdb_git_hash system variable.

• Percona Server doesn't provide either.

RocksDB Version

• Facebook's branch uses RocksDB 5.10.0 (the version number can be found in include/rocksdb/version.h)

commit ba295cda29daee3ffe58549542804efdfd969784
Author: Andrew Kryczka <andrewkr@fb.com>
Date:   Fri Jan 12 11:03:55 2018 -0800

• MariaDB currently uses 5.8.0

commit 9a970c81af9807071bd690f4c808c5045866291a
Author: Yi Wu <yiwu@fb.com>
Date:   Wed Sep 13 17:21:35 2017 -0700

• Percona Server uses 5.8.0

commit ab0542f5ec6e7c7e405267eaa2e2a603a77d570b
Author: Maysam Yabandeh <myabandeh@fb.com>
Date:   Fri Sep 29 07:55:22 2017 -0700

Binlog Position in information_schema.rocksdb_global_info

• FB branch provides information_schema.rocksdb_global_info type=BINLOG, NAME={FILE, POS, GTID}.

• Percona Server doesn't provide it.

• MariaDB doesn't provide it.

One use of that information is to take the output of myrocks_hotbackup and make it a new master.

Gap Lock Detector

• FB branch has a "Gap Lock Detector" feature. It is at the SQL layer. It can be controlled with gap_lock_XXX variables and is disabled by default (gap-lock-raise-error=false, gap-lock-write-lock=false).

• Percona Server has gap lock checking ON but doesn't seem to have any way to control it? Queries that use Gap Lock on MyRocks fail with an error like this:

INSERT INTO tbl2 SELECT * FROM tbl1;
ERROR 1105 (HY000): Using Gap Lock without full unique key in multi-table or multi-statement transactions
is not allowed. You need to either rewrite queries to use all unique key columns in WHERE equal conditions,
or rewrite to single-table, single-statement transaction.  Query: insert into tbl2 select * from tbl1

• MariaDB doesn't include the Gap Lock Detector.

Generated Columns

• Both MariaDB and Percona Server support generated columns, but neither one supports them for the MyRocks storage engine (attempts to create a table will produce an error).

• Invisible columns in MariaDB 10.3 are supported (as they are an SQL layer feature).

rpl_skip_tx_api

Facebook's branch has a performance feature for replication slaves, rpl_skip_tx_api. It is not available in MariaDB or in Percona Server.

Details

The above comparison was made using

• FB/MySQL 5.6.35

• Percona Server 5.7.20-19-log

• MariaDB 10.2.13 (MyRocks is beta)

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks is a storage engine that adds the RocksDB database to MariaDB. RocksDB is an LSM database with a great compression ratio that is optimized for flash storage.

The storage engine must be installed before it can be used.

Installing the Plugin's Package

The MyRocks storage engine's shared library is included in MariaDB packages as the ha_rocksdb.so or ha_rocksdb.dll shared library on systems where it can be built.

Installing on Linux

The MyRocks storage engine is included in binary tarballs on Linux.

Installing with a Package Manager

The MyRocks storage engine can also be installed via a package manager on Linux. In order to do so, your system needs to be configured to install from one of the MariaDB repositories.

You can configure your package manager to install it from MariaDB Corporation's MariaDB Package Repository by using the MariaDB Package Repository setup script.

You can also configure your package manager to install it from MariaDB Foundation's MariaDB Repository by using the MariaDB Repository Configuration Tool.

Installing with yum/dnf

On RHEL, CentOS, Fedora, and other similar Linux distributions, it is highly recommended to install the relevant RPM package from MariaDB's repository using yum or dnf. Starting with RHEL 8 and Fedora 22, yum has been replaced by dnf, which is the next major version of yum. However, yum commands still work on many systems that use dnf:

sudo yum install MariaDB-rocksdb-engine

Installing with apt-get

On Debian, Ubuntu, and other similar Linux distributions, it is highly recommended to install the relevant DEB package from MariaDB's repository using apt-get:

sudo apt-get install mariadb-plugin-rocksdb

Installing with zypper

On SLES, OpenSUSE, and other similar Linux distributions, it is highly recommended to install the relevant RPM package from MariaDB's repository using zypper:

sudo zypper install MariaDB-rocksdb-engine

Installing on Windows

The MyRocks storage engine is included in MSI and ZIP packages on Windows.

Installing the Plugin

Once the shared library is in place, the plugin is not actually installed by MariaDB by default. There are two methods that can be used to install the plugin with MariaDB.

The first method can be used to install the plugin without restarting the server. You can install the plugin dynamically by executing INSTALL SONAME or INSTALL PLUGIN:

INSTALL SONAME 'ha_rocksdb';

The second method can be used to tell the server to load the plugin when it starts up. The plugin can be installed this way by providing the --plugin-load or the --plugin-load-add options. This can be specified as a command-line argument to mysqld or it can be specified in a relevant server option group in an option file:

[mariadb]
...
plugin_load_add = ha_rocksdb

Note: When installed with a package manager, an option file that contains the --plugin-load-add option may also be installed. The RPM package installs it as /etc/my.cnf.d/rocksdb.cnf, and the DEB package installs it as /etc/mysql/mariadb.conf.d/rocksdb.cnf

Uninstalling the Plugin

You can uninstall the plugin dynamically by executing UNINSTALL SONAME or UNINSTALL PLUGIN:

UNINSTALL SONAME 'ha_rocksdb';

If you installed the plugin by providing the --plugin-load or the --plugin-load-add options in a relevant server option group in an option file, then those options should be removed to prevent the plugin from being loaded the next time the server is restarted.

Verifying the Installation

After installing MyRocks you will see RocksDB in the list of plugins:

SHOW PLUGINS;
+-------------------------------+----------+--------------------+---------------+---------+
| Name                          | Status   | Type               | Library       | License |
+-------------------------------+----------+--------------------+---------------+---------+
...
| ROCKSDB                       | ACTIVE   | STORAGE ENGINE     | ha_rocksdb.so | GPL     |
| ROCKSDB_CFSTATS               | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_DBSTATS               | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_PERF_CONTEXT          | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_PERF_CONTEXT_GLOBAL   | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_CF_OPTIONS            | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_COMPACTION_STATS      | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_GLOBAL_INFO           | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_DDL                   | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_INDEX_FILE_MAP        | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_LOCKS                 | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
| ROCKSDB_TRX                   | ACTIVE   | INFORMATION SCHEMA | ha_rocksdb.so | GPL     |
...
+-------------------------------+----------+--------------------+---------------+---------+

Compression

Supported compression types are listed in the rocksdb_supported_compression_types variable:

SHOW VARIABLES LIKE 'rocksdb_supported_compression_types';
+-------------------------------------+-------------+
| Variable_name                       | Value       |
+-------------------------------------+-------------+
| rocksdb_supported_compression_types | Snappy,Zlib |
+-------------------------------------+-------------+

See MyRocks and Data Compression for more.

System and Status Variables

All MyRocks system variables and status variables are prefaced with "rocksdb", so you can query them with, for example:

SHOW VARIABLES LIKE 'rocksdb%';
SHOW STATUS LIKE 'rocksdb%';

_{This page is licensed: CC BY-SA / Gnu FDL}

Being a write-optimized storage engine, MyRocks has special ways to load data much faster than normal INSERTs would.

See

• ; the section about "Migrating from InnoDB to MyRocks in production" has some clues.

• Data-Loading covers the topic in greater detail.

Note When one loads data with rocksdb_bulk_load=1 and the data conflicts with the data already in the database, one may get non-trivial errors, for example:

ERROR 1105 (HY000): [./.rocksdb/test.t1_PRIMARY_2_0.bulk_load.tmp] bulk load error: 
  Invalid argument: External file requires flush

_{This page is licensed: CC BY-SA / Gnu FDL}

Bloom filters are used to reduce read amplification. Bloom filters can be set on a per-column family basis (see myrocks-column-families).

Bloom Filter Parameters

• How many bits to use

• whole_key_filtering=true/false

• Whether the bloom filter is for the entire key or for the prefix. In case of a prefix, you need to look at the index definition and compute the desired prefix length.

Computing Prefix Length

• It's 4 bytes for index_nr

• Then, for fixed-size columns (integer, date[time], decimal) it is key_length as shown by EXPLAIN. For VARCHAR columns, determining the length is tricky (It depends on the values stored in the table. Note that MyRocks encodes VARCHARs with "Variable-Length Space-Padded Encoding" format).

Configuring Bloom Filter

To enable 10-bit bloom filter for 8-byte prefix length for column family "cf1", put this into my.cnf:

rocksdb_override_cf_options='cf1={block_based_table_factory={filter_policy=bloomfilter:10:false;whole_key_filtering=0;};prefix_extractor=capped:8};'

and restart the server.

Check if the column family actually uses the bloom filter:

SELECT * 
FROM information_schema.rocksdb_cf_options 
WHERE 
  cf_name='cf1' AND
  option_type IN ('TABLE_FACTORY::FILTER_POLICY','PREFIX_EXTRACTOR');

+---------+------------------------------+----------------------------+
| CF_NAME | OPTION_TYPE                  | VALUE                      |
+---------+------------------------------+----------------------------+
| cf1     | PREFIX_EXTRACTOR             | rocksdb.CappedPrefix.8     |
| cf1     | TABLE_FACTORY::FILTER_POLICY | rocksdb.BuiltinBloomFilter |
+---------+------------------------------+----------------------------+

Checking if Bloom Filter is Useful

Watch these status variables:

SHOW status LIKE '%bloom%';
+-------------------------------------+-------+
| Variable_name                       | Value |
+-------------------------------------+-------+
| Rocksdb_bloom_filter_prefix_checked | 1     |
| Rocksdb_bloom_filter_prefix_useful  | 0     |
| Rocksdb_bloom_filter_useful         | 0     |
+-------------------------------------+-------+

Other useful variables are:

• rocksdb_force_flush_memtable_now - bloom filter is only used when reading data from disk. If you are doing testing, flush the data to disk first.

• rocksdb_skip_bloom_filter_on_read - skip using the bloom filter (default is FALSE).

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks supports the CHECK TABLE command.

The command will do a number of checks to verify that the table data is self-consistent.

The details about the errors are printed into the error log. If log_warnings > 2, the error log will also have some informational messages which can help with troubleshooting.

Besides this, RocksDB has its own (low-level) log in #rocksdb/LOG file.

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks supports several compression algorithms.

Supported Compression Algorithms

Supported compression algorithms can be checked like so:

SHOW variables LIKE 'rocksdb%compress%';
+-------------------------------------+------------------------------------+
| Variable_name                       | Value                              |
+-------------------------------------+------------------------------------+
| rocksdb_supported_compression_types | Snappy,Zlib,LZ4,LZ4HC,ZSTDNotFinal |
+-------------------------------------+------------------------------------+

Another way to make the check is to look into #rocksdb/LOG file in the data directory. It should have lines like:

2019/04/12-14:08:23.869919 7f839188b540 Compression algorithms supported:
2019/04/12-14:08:23.869920 7f839188b540         kZSTDNotFinalCompression supported: 1
2019/04/12-14:08:23.869922 7f839188b540         kZSTD supported: 1
2019/04/12-14:08:23.869923 7f839188b540         kXpressCompression supported: 0
2019/04/12-14:08:23.869924 7f839188b540         kLZ4HCCompression supported: 1
2019/04/12-14:08:23.869924 7f839188b540         kLZ4Compression supported: 1
2019/04/12-14:08:23.869925 7f839188b540         kBZip2Compression supported: 0
2019/04/12-14:08:23.869926 7f839188b540         kZlibCompression supported: 1
2019/04/12-14:08:23.869927 7f839188b540         kSnappyCompression supported: 1

Compression Settings

Compression is set on a per-Column Family basis. See MyRocks Column Families.

Checking Compression Settings

To check current compression settings for a column family one can use a query like so:

SELECT * FROM information_schema.rocksdb_cf_options 
WHERE option_type LIKE '%ompression%' AND cf_name='DEFAULT';

The output are like:

+---------+-----------------------------------------+---------------------------+
| CF_NAME | OPTION_TYPE                             | VALUE                     |
+---------+-----------------------------------------+---------------------------+
| default | COMPRESSION_TYPE                        | kSnappyCompression        |
| default | COMPRESSION_PER_LEVEL                   | NUL                       |
| default | COMPRESSION_OPTS                        | -14:32767:0               |
| default | BOTTOMMOST_COMPRESSION                  | kDisableCompressionOption |
| default | TABLE_FACTORY::VERIFY_COMPRESSION       | 0                         |
| default | TABLE_FACTORY::ENABLE_INDEX_COMPRESSION | 1                         |
+---------+-----------------------------------------+---------------------------+

Current column family settings are used for the new SST files.

Modifying Compression Settings

Compression settings are not dynamic parameters, one cannot change them by setting rocksdb_update_cf_options.

The procedure to change compression settings is as follows:

• Edit my.cnf to set rocksdb_override_cf_options.

Example:

rocksdb-override-cf-options='cf1={compression=kZSTD;bottommost_compression=kZSTD;}'

• Restart the server.

The data will not be re-compressed immediately. However, all new SST files will use the new compression settings, so as data gets inserted/updated the column family will gradually start using the new option.

Caveat: Syntax Errors

Please note that rocksdb-override-cf-options syntax is quite strict. Any typos will result in the parse error, and MyRocks plugin will not be loaded. Depending on your configuration, the server may still start. If it does start, you can use this command to check if the plugin is loaded:

SELECT * FROM information_schema.plugins WHERE plugin_name='ROCKSDB'

(note that you need the "ROCKSDB" plugin. Other auxiliary plugins like "ROCKSDB_TRX" might still get loaded).

Another way is to detect the error is check the error log. When option parsing fails, it will contain messages like so:

2019-04-16 11:07:57 140283675678016 [Warning] Invalid cf config for cf1 in override options (options: cf1={compression=kLZ4Compression;bottommost_compression=kZSTDCompression;})
2019-04-16 11:07:57 140283675678016 [ERROR] RocksDB: Failed to initialize CF options map.
2019-04-16 11:07:57 140283675678016 [ERROR] Plugin 'ROCKSDB' init function returned error.
2019-04-16 11:07:57 140283675678016 [ERROR] Plugin 'ROCKSDB' registration as a STORAGE ENGINE failed.

Checking How the Data is Compressed

A query to check what compression is used in the SST files that store the data for a given table (test.t1):

SELECT
  SP.sst_name, SP.compression_algo
FROM
  information_schema.rocksdb_sst_props SP,
  information_schema.rocksdb_ddl D,
  information_schema.rocksdb_index_file_map IFM
WHERE
  D.table_schema='test' AND D.table_name='t1' AND
  D.index_number= IFM.index_number AND
  IFM.sst_name=SP.sst_name;

Example output:

+------------+------------------+
| sst_name   | compression_algo |
+------------+------------------+
| 000028.sst | Snappy           |
| 000028.sst | Snappy           |
| 000026.sst | Snappy           |
| 000026.sst | Snappy           |
+------------+------------------+

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks supports group commit with the binary log (MDEV-11934).

Counter Values to Watch

(The following is only necessary if you are studying MyRocks internals)

MariaDB's group commit counters are:

Binlog_commits - how many transactions were written to the binary log

Binlog_group_commits - how many group commits happened. (e.g. if each group had two transactions, this is twice as small as Binlog_commits)

On the RocksDB side, there is one relevant counter:Rocksdb_wal_synced - How many times RocksDB's WAL file was synced. (TODO: this is after group commit happened, right?)

On the Value of rocksdb_wal_group_syncs

FB/MySQL-5.6 has a rocksdb_wal_group_syncs counter (The counter is provided by MyRocks, it is not a view of a RocksDB counter). It is increased in rocksdb_flush_wal() when doing the rdb->FlushWAL() call.

rocksdb_flush_wal() is called by MySQL's Group Commit when it wants to make the effect of several rocksdb_prepare() calls persistent.

So, the value of rocksdb_wal_group_syncs in FB/MySQL-5.6 is similar to Binlog_group_commits in MariaDB.

MariaDB doesn't have that call, each rocksdb_prepare() call takes care of being persistent on its own.

Because of that, rocksdb_wal_group_syncs is zero for MariaDB. (Currently, it is only incremented when the binlog is rotated).

Examples

So for a workload with concurrency=50, n_queries=10K, one gets

• Binlog_commits=10K

• Binlog_group_commits=794

• Rocksdb_wal_synced=8362

This is on a RAM disk

For a workload with concurrency=50, n_queries=10K, rotating laptop hdd, one gets

• Binlog_commits= 10K

• Binlog_group_commits=1403

• Rocksdb_wal_synced=400

The test took 38 seconds, Number of syncs was 1400+400=1800, which gives 45 syncs/sec which looks normal for this slow rotating desktop hdd.

Note that the WAL was synced fewer times than there were binlog commit groups (?)

_{This page is licensed: CC BY-SA / Gnu FDL}

This article is about MyRocks and index-only scans on secondary indexes. It applies to MariaDB's MyRocks, Facebook's MyRocks, and other variants.

Secondary Keys Only

The primary key in MyRocks is always the clustered key, that is, the index record is THE table record and so it's not possible to do "index only" because there isn't anything that is not in the primary key's (Key,Value) pair.

Secondary keys may or may not support index-only scans, depending on the datatypes of the columns that the query is trying to read.

Background: Mem-Comparable Keys

MyRocks indexes store "mem-comparable keys" (that is, the key values are compared with memcmp). For some datatypes, it is easily possible to convert between the column value and its mem-comparable form, while for others the conversion is one-way.

For example, in case-insensitive collations capital and regular letters are considered identical, i.e. 'c' ='C'. For some datatypes, MyRocks stores some extra data which allows it to restore the original value back. (For the latin1_general_ci collation and character 'c', for example, it will store one bit which says whether the original value was a small 'c' or a capital letter 'C'). This doesn't work for all datatypes, though.

Index-Only Support for Various Datatypes

Index-only scans are supported for numeric and date/time datatypes. For CHAR and VAR[CHAR], it depends on which collation is used, see below for details.

Index-only scans are currently not supported for less frequently used datatypes, like

• BIT(n)

• SET(...)

• ENUM(...) It is actually possible to add support for those, feel free to write a patch or at least make a case why a particular datatype is important

Index-Only Support for Various Collations

As far as Index-only support is concerned, MyRocks distinguishes three kinds of collations:

1. Binary (Reversible) Collations

These are binary, latin1_bin, and utf8_bin.

For these collations, it is possible to convert a value back from its mem-comparable form. Hence, one can restore the original value back from its index record, and index-only scans are supported.

2. Restorable Collations

These are collations where one can store some extra information which helps to restore the original value.

Criteria (from storage/rocksdb/rdb_datadic.cc, rdb_is_collation_supported()) are:

• The charset should use 1-byte characters (so, unicode-based collations are not included)

• strxfrm(1 byte) = {one 1-byte weight value always}

• no binary sorting

• PAD attribute

The examples are: latin1_general_ci, latin1_general_cs, latin1_swedish_ci, etc.

Index-only scans are supported for these collations.

3. All Other Collations

For these collations, there is no known way to restore the value from its mem-comparable form, and so index-only scans are not supported.

MyRocks needs to fetch the clustered PK record to get the field value.

Covering Secondary Key Lookups for VARCHARs

TODO: there is also this optimization:

303f349c95848e92b5b27b44f0e57194100eb0997e7

document it.

_{This page is licensed: CC BY-SA / Gnu FDL}

Details about how MyRocks works with replication.

MyRocks and Statement-Based Replication

Statement-based replication (SBR) works as follows: SQL statements are executed on the master (possibly concurrently). They are written into the binlog (this fixes their ordering, "a serialization"). The slave then reads the binlog and executes the statements in their binlog order.

In order to prevent data drift, serial execution of statements on the slave must have the same effect as concurrent execution of these statements on the master. In other words, transaction isolation on the master must be close to SERIALIZABLE transaction isolation level (This is not a strict mathematical proof but shows the idea).

InnoDB achieves this by (almost) supporting SERIALIZABLE transactional isolation level. It does so by supporting "Gap Locks". MyRocks doesn't support SERIALIZABLE isolation, and it doesn't support gap locks.

Because of that, generally one cannot use MyRocks and statement-based replication.

Updating a MyRocks table while having SBR on, will result in an error as follow:

ERROR 4056 (HY000): Can't execute updates on master with binlog_format != ROW.

Can One Still Use SBR with MyRocks?

Yes. In many cases, database applications run a restricted set of SQL statements, and it's possible to prove that lack of Gap Lock support is not a problem and data skew will not occur.

In that case, one can set @@rocksdb_unsafe_for_binlog=1 and MyRocks will work with SBR. The user is however responsible for making sure their queries are not causing a data skew.

Differences From Upstream MyRocks

MyRocks upstream (that is, Facebook's MySQL branch) has a number of unique replication enhancements. These are available in upstream's version of MyRocks but not in MariaDB's version of MyRocks.

• Read-Free Replication (see Read-Free-Replication) TODO

• <<unique_check_lag_threshold>>. This is FB/MySQL-5.6 feature where unique checks are disabled if replication lag exceeds a certain threshold.

• <<slave_gtid_info=OPTIMIZED>>. This is said to be:

<<quote>>
"Whether SQL threads update mysql.slave_gtid_info table. If this value "
"is OPTIMIZED, updating the table is done inside storage engines to "
"avoid MySQL layer's performance overhead",
<</quote>>

_{This page is licensed: CC BY-SA / Gnu FDL}

FB/MySQL has added new syntax:

START TRANSACTION WITH CONSISTENT ROCKSDB|INNODB SNAPSHOT;

The statement returns the binlog coordinates pointing at the snapshot.

MariaDB (and Percona Server) support extension to the regular

START TRANSACTION WITH CONSISTENT SNAPSHOT;

syntax as documented in Enhancements for START TRANSACTION WITH CONSISTENT SNAPSHOT.

After issuing the statement, one can examine the binlog_snapshot_file and binlog_snapshot_position status variables to see the binlog position that corresponds to the snapshot.

See Also

• START TRANSACTION

• Enhancements for START TRANSACTION WITH CONSISTENT SNAPSHOT

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks stores data in column families. These are similar to tablespaces. By default, the data is stored in the default column family.

One can specify which column family the data goes to by using index comments:

INDEX index_name(col1, col2, ...) COMMENT 'column_family_name'

If the column name starts with rev:, the column family is reverse-ordered.

Reasons for Column Families

Storage parameters like

• Bloom filter settings

• Compression settings

• Whether the data is stored in reverse order

are specified on a per-column family basis.

Creating a Column Family

When creating a table or index, you can specify the name of the column family for it. If the column family doesn't exist, it are automatically created.

Dropping a Column Family

There is currently no way to drop a column family. RocksDB supports this internally but MyRocks doesn't provide any way to do it.

Setting Column Family Parameters

Use these variables:

• rocksdb_default_cf_options - a my.cnf parameter specifying default options for all column families.

• rocksdb_override_cf_options - a my.cnf parameter specifying per-column family option overrides.

• rocksdb_update_cf_options - a dynamically-settable variable which allows to change parameters online. Not all parameters can be changed.

rocksdb_override_cf_options

This parameter allows one to override column family options for specific column families. Here is an example of how to set option1=value1 and option2=value for column family cf1, and option3=value3 for column family cf3:

rocksdb_override_cf_options='cf1={option1=value1;option2=value2};cf2={option3=value3}'

One can check the contents of INFORMATION_SCHEMA.ROCKSDB_CF_OPTIONS to see what options are available.

Options that are frequently configured are:

• Data compression. See myrocks-and-data-compression.

• Bloom Filters. See myrocks-and-bloom-filters.

Examining Column Family Parameters

See the INFORMATION_SCHEMA.ROCKSDB_CF_OPTIONS table.

_{This page is licensed: CC BY-SA / Gnu FDL}

MyRocks exposes its performance metrics through several interfaces:

• Status variables

• SHOW ENGINE ROCKSDB STATUS

• RocksDB's perf context

the contents slightly overlap, but each source has its own unique information, so be sure to check all three.

Status Variables

Check the output of

SHOW STATUS LIKE 'Rocksdb%'

See MyRocks Status Variables for more information.

SHOW ENGINE ROCKSDB STATUS

This produces a lot of information.

One particularly interesting part is compaction statistics. It shows the amount of data on each SST level and other details:

*************************** 4. row ***************************
  Type: CF_COMPACTION
  Name: default
Status: 
** Compaction Stats [default] **
Level    Files   Size     Score Read(GB)  Rn(GB) Rnp1(GB) Write(GB) Wnew(GB) Moved(GB) W-Amp Rd(MB/s) Wr(MB/s) Comp(sec) Comp(cnt) Avg(sec) KeyIn KeyDrop
----------------------------------------------------------------------------------------------------------------------------------------------------------
  L0      3/0   30.16 MB   1.0      0.0     0.0      0.0      11.9     11.9       0.0   1.0      0.0     76.6       159       632    0.251       0      0
  L1      5/0   247.54 MB   1.0      0.7     0.2      0.5       0.5      0.0      11.6   2.6     58.5     44.1        12         4    2.926     30M    10M
  L2    112/0    2.41 GB   1.0      0.6     0.0      0.6       0.5     -0.1      11.4  43.4     55.2     45.9        11         1   10.827     21M  3588K
  L3    466/0    8.91 GB   0.4      0.0     0.0      0.0       0.0      0.0       8.9   0.0      0.0      0.0         0         0    0.000       0      0
 Sum    586/0   11.59 GB   0.0      1.3     0.2      1.0      12.8     11.8      32.0   1.1      7.1     72.6       181       637    0.284     52M    13M
 Int      0/0    0.00 KB   0.0      0.9     0.1      0.8       0.8      0.0       0.1  20.5     48.4     45.3        19         6    3.133     33M  3588K

Performance Context

RocksDB has an internal mechanism called "perf context". The counter values are exposed through two tables:

• INFORMATION_SCHEMA.ROCKSDB_PERF_CONTEXT_GLOBAL - global counters

• INFORMATION_SCHEMA.ROCKSDB_PERF_CONTEXT - Per-table/partition counters

By default statistics are NOT collected. One needs to set rocksdb_perf_context_level to some value (e.g. 3) to enable collection.

_{This page is licensed: CC BY-SA / Gnu FDL}

This article describes how MyRocks storage engine provides statistics to the query optimizer.

There are three kinds of statistics:

• Table statistics (number of rows in the table, average row size)

• Index cardinality (how distinct values are in the index)

• records-in-range estimates (how many rows are in a certain range "const1 < tbl.key < const2".

How MyRocks computes statistics

MyRocks (actually RocksDB) uses LSM files which are written once and never updated. When an LSM file is written, MyRocks will compute index cardinalities and number-of-rows for the data in the file. (The file generally has rows, index records and/or tombstones for multiple tables/indexes).

For performance reasons, statistics are computed based on a fraction of rows in the LSM file. The percentage of rows used is controlled by rocksdb_table_stats_sampling_pct; the default value is 10%.

Before the data is dumped into LSM file, it is stored in the MemTable. MemTable doesn't allow computing index cardinalities, but it can provide an approximate number of rows in the table. Use of MemTable data for statistics is controlled by rocksdb_force_compute_memtable_stats; the default value is ON.

Are index statistics predictable?

Those who create/run MTR tests, need to know whether EXPLAIN output is deterministic. For MyRocks tables, the answer is NO (just like for InnoDB).

Statistics are computed using sampling and GetApproximateMemTableStats() which means that the #rows column in the EXPLAIN output may vary slightly.

Records-in-range estimates

MyRocks uses RocksDB's GetApproximateSizes() call to produce an estimate for the number of rows in the certain range. The data in MemTable is also taken into account by issuing a GetApproximateMemTableStats call.

ANALYZE TABLE

ANALYZE TABLE will possibly flush the MemTable (depending on the rocksdb_flush_memtable_on_analyze and rocksdb_pause_background_work settings).

After that, it will re-read statistics from the SST files and re-compute the summary numbers (TODO: and if the data was already on disk, the result should not be different from the one we had before ANALYZE?)

Debugging helper variables

There are a few variables that will cause MyRocks to report certain pre-defined estimate numbers to the optimizer:

• @@rocksdb_records_in_range - if not 0, report that any range has this many rows

• @@rocksdb_force_index_records_in_range - if not 0, and FORCE INDEX hint is used, report that any range has this many rows.

• @@rocksdb_debug_optimizer_n_rows - if not 0, report that any MyRocks table has this many rows.

_{This page is licensed: CC BY-SA / Gnu FDL}

TODO:

• MyRocks uses snapshot isolation

• Support do READ-COMMITTED and REPEATABLE-READ

• SERIALIZABLE is not supported

  • There is no "Gap Locking" which makes Statement Based Replication unsafe (See MyRocks and Replication).

_{This page is licensed: CC BY-SA / Gnu FDL}

This page documents status variables related to the MyRocks storage engine. See Server Status Variables for a complete list of status variables that can be viewed with SHOW STATUS.

See also the Full list of MariaDB options, system and status variables.

Rocksdb_block_cache_add

• Description: Number of blocks added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_add_failures

• Description: Number of failures when adding blocks to Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_bytes_read

• Description: Bytes read from Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_bytes_write

• Description: Bytes written to Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_data_add

• Description: Number of data blocks added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_data_bytes_insert

• Description: Bytes added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_data_hit

• Description: Number of hits when accessing the data block from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_data_miss

• Description: Number of misses when accessing the data block from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_filter_add

• Description: Number of bloom filter blocks added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_filter_bytes_evict

• Description: Bytes of bloom filter blocks evicted from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_filter_bytes_insert

• Description: Bytes of bloom filter blocks added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_filter_hit

• Description: Number of hits when accessing the filter block from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_filter_miss

• Description: Number of misses when accessing the filter block from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_hit

• Description: Total number of hits for the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_index_add

• Description: Number of index blocks added to Block Cache index.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_index_bytes_evict

• Description: Bytes of index blocks evicted from the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_index_bytes_insert

• Description: Bytes of index blocks added to the Block Cache.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_block_cache_index_hit

• Description: Number of hits for the Block Cache index.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_index_miss

• Description: Number of misses for the Block Cache index.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cache_miss

• Description: Total number of misses for the Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cachecompressed_hit

• Description: Number of hits for the compressed Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_block_cachecompressed_miss

• Description: Number of misses for the compressed Block Cache.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_bloom_filter_full_positive

• Description:

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.18, MariaDB 10.3.10

Rocksdb_bloom_filter_full_true_positive

• Description:

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.18, MariaDB 10.3.10

Rocksdb_bloom_filter_prefix_checked

• Description: Number of times the Bloom Filter checked before creating an iterator on a file.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_bloom_filter_prefix_useful

• Description: Number of times the Bloom Filter check used to avoid creating an iterator on a file.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_bloom_filter_useful

• Description: Number of times the Bloom Filter used instead of reading form file.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_bytes_read

• Description: Total number of uncompressed bytes read from memtables, cache or table files.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_bytes_written

• Description: Total number of uncompressed bytes written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_compact_read_bytes

• Description: Number of bytes read during compaction.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_compact_write_bytes

• Description: Number of bytes written during compaction.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_compaction_key_drop_new

• Description: Number of keys dropped during compaction due their being overwritten by new values.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_compaction_key_drop_obsolete

• Description: Number of keys dropped during compaction due to their being obsolete.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_compaction_key_drop_user

• Description: Number of keys dropped during compaction due to user compaction.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_covered_secondary_key_lookups

• Description: Incremented when avoiding reading a record via a keyread. This indicates lookups that were performed via a secondary index containing a field that is only a prefix of the VARCHAR column, and that could return all requested fields directly from the secondary index.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_flush_write_bytes

• Description: Number of bytes written during flush.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_get_hit_l0

• Description: Number of times reads got data from the L0 compaction layer.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_get_hit_l1

• Description: Number of times reads got data from the L1 compaction layer.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_get_hit_l2_and_up

• Description: Number of times reads got data from the L2 and up compaction layer.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_getupdatessince_calls

• Description: Number of calls to the GetUpdatesSince function. You may find this useful when monitoring refreshes of the transaction log.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_iter_bytes_read

• Description: Total uncompressed bytes read from an iterator, including the size of both key and value.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_l0_num_files_stall_micros

• Description: Shows how long in microseconds throttled due to too mnay files in L0.

• Scope: Global, Session

• Data Type: numeric

• Removed: MariaDB 10.3.1, MariaDB 10.2.8

Rocksdb_l0_slowdown_micros

• Description: Total time spent waiting in microseconds while performing L0-L1 compactions.

• Scope: Global, Session

• Data Type: numeric

• Removed: MariaDB 10.3.1, MariaDB 10.2.8

Rocksdb_manual_compactions_processed

• Description:

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.18, MariaDB 10.3.10

Rocksdb_manual_compactions_running

• Description:

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.18, MariaDB 10.3.10

Rocksdb_memtable_compaction_micros

• Description:

• Scope: Global, Session

• Data Type: numeric

• Removed: MariaDB 10.3.1, MariaDB 10.2.8

Rocksdb_memtable_hit

• Description: Number of memtable hits.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_memtable_miss

• Description: Number of memtable misses.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_memtable_total

• Description: Memory used, in bytes, of all memtables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_memtable_unflushed

• Description: Memory used, in bytes, of all unflushed memtables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_no_file_closes

• Description: Number of times files were closed.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_no_file_errors

• Description: Number of errors encountered while trying to read data from an SST file.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_no_file_opens

• Description: Number of times files were opened.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_num_iterators

• Description: Number of iterators currently open.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_block_not_compressed

• Description: Number of uncompressed blocks.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_db_next

• Description: Number of next calls.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_db_next_found

• Description: Number of next calls that returned data.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_db_prev

• Description: Number of prev calls.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_db_prev_found

• Description: Number of prev calls that returned data.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_db_seek

• Description: Number of seek calls.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_db_seek_found

• Description: Number of seek calls that returned data.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_number_deletes_filtered

• Description: Number of deleted records were not written to storage due to a nonexistent key.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_keys_read

• Description: Number of keys have been read.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_keys_updated

• Description: Number of keys have been updated.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_keys_written

• Description: Number of keys have been written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_merge_failures

• Description: Number of failures encountered while performing merge operator actions.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_multiget_bytes_read

• Description: Number of bytes read during RocksDB MultiGet() calls.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_multiget_get

• Description: Number of RocksDB MultiGet() requests made.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_multiget_keys_read

• Description: Number of keys read through RocksDB MultiGet() calls.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_reseeks_iteration

• Description: Number of reseeks that have occurred inside an iteration that skipped over a large number of keys with the same user key.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_sst_entry_delete

• Description: Number of delete markers written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_sst_entry_merge

• Description: Number of merge keys written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_sst_entry_other

• Description: Number of keys written that are not delete, merge or put keys.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_sst_entry_put

• Description: Number of put keys written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_sst_entry_singledelete

• Description: Number of single-delete keys written.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_superversion_acquires

• Description: Number of times the superversion structure acquired. This is useful when tracking files for the database.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_superversion_cleanups

• Description: Number of times the superversion structure performed cleanups.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_number_superversion_releases

• Description: Number of times the superversion structure released.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_queries_point

• Description: Number of single-row queries.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_queries_range

• Description: Number of multi-row queries.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_row_lock_deadlocks

• Description: Number of deadlocks.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_row_lock_wait_timeouts

• Description: Number of row lock wait timeouts.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_rows_deleted

• Description: Number of rows deleted.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_rows_deleted_blind

• Description:

• Scope: Global, Session

• Data Type: numeric

Rocksdb_rows_expired

• Description: Number of expired rows.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_rows_filtered

• Description: Number of TTL filtered rows.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.15, MariaDB 10.3.7

Rocksdb_rows_inserted

• Description: Number of rows inserted.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_rows_read

• Description: Number of rows read.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_rows_updated

• Description: Number of rows updated.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_snapshot_conflict_errors

• Description: Number of snapshot conflict errors that have occurred during transactions that forced a rollback.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_l0_file_count_limit_slowdowns

• Description: Write slowdowns due to L0 being near to full.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_l0_file_count_limit_stops

• Description: Write stops due to L0 being to full.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_locked_l0_file_count_limit_slowdowns

• Description: Write slowdowns due to L0 being near to full and L0 compaction in progress.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_locked_l0_file_count_limit_stops

• Description: Write stops due to L0 being full and L0 compaction in progress.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_memtable_limit_slowdowns

• Description: Write slowdowns due to approaching maximum permitted number of memtables.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.10, MariaDB 10.3.3

Rocksdb_stall_memtable_limit_stops

• Description: * Description: Write stops due to reaching maximum permitted number of memtables.

• Scope: Global, Session

• Data Type: numeric

• Introduced: MariaDB 10.2.10, MariaDB 10.3.3

Rocksdb_stall_micros

• Description: Time in microseconds that the writer had to wait for the compaction or flush to complete.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_pending_compaction_limit_slowdowns

• Description: Write slowdowns due to nearing the limit for the maximum number of pending compaction bytes.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_pending_compaction_limit_stops

• Description: Write stops due to reaching the limit for the maximum number of pending compaction bytes.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_total_slowdowns

• Description: Total number of write slowdowns.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_stall_total_stops

• Description: Total number of write stops.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_system_rows_deleted

• Description: Number of rows deleted from system tables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_system_rows_inserted

• Description: Number of rows inserted into system tables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_system_rows_read

• Description: Number of rows read from system tables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_system_rows_updated

• Description: Number of rows updated for system tables.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_wal_bytes

• Description: Number of bytes written to WAL.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_wal_group_syncs

• Description: Number of group commit WAL file syncs have occurred. This is provided by MyRocks and is not a view of a RocksDB counter. Increased in rocksdb_flush_wal() when doing the rdb->FlushWAL() call.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_wal_synced

• Description: Number of syncs made on RocksDB WAL file.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_write_other

• Description: Number of writes processed by a thread other than the requesting thread.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_write_self

• Description: Number of writes processed by requesting thread.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_write_timedout

• Description: Number of writes that timed out.

• Scope: Global, Session

• Data Type: numeric

Rocksdb_write_wal

• Description: Number of write calls that requested WAL.

• Scope: Global, Session

• Data Type: numeric

_{This page is licensed: CC BY-SA / Gnu FDL}

This page documents system variables related to the MyRocks storage engine. See Server System Variables for a complete list of system variables and instructions on setting them.

See also the Full list of MariaDB options, system and status variables.

rocksdb_access_hint_on_compaction_start

• Description: DBOptions::access_hint_on_compaction_start for RocksDB. Specifies the file access pattern, applied to all input files, once a compaction starts.

• Command line: --rocksdb-access-hint-on-compaction-start=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 1

• Range: 0 to 3

rocksdb_advise_random_on_open

• Description: DBOptions::advise_random_on_open for RocksDB.

• Command line: --rocksdb-advise-random-on-open={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_allow_concurrent_memtable_write

• Description: DBOptions::allow_concurrent_memtable_write for RocksDB.

• Command line: --rocksdb-allow-concurrent-memtable-write={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_allow_mmap_reads

• Description: DBOptions::allow_mmap_reads for RocksDB

• Command line: --rocksdb-allow-mmap-reads={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_allow_mmap_writes

• Description: DBOptions::allow_mmap_writes for RocksDB

• Command line: --rocksdb-allow-mmap-writes={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_allow_to_start_after_corruption

• Description: Allow server still to start successfully even if RocksDB corruption is detected.

• Command line: --rocksdb-allow-to-start-after-corruption={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_background_sync

• Description: Turns on background syncs for RocksDB

• Command line: --rocksdb-background-sync={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

• Removed: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_base_background_compactions

• Description: DBOptions::base_background_compactions for RocksDB

• Command line: --rocksdb-base-background-compactions=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 1

• Range: -1 to 64

• Removed: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_blind_delete_primary_key

• Description: Deleting rows by primary key lookup, without reading rows (Blind Deletes). Blind delete is disabled if the table has secondary key.

• Command line: --rocksdb-blind-delete-primary-key={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_block_cache_size

• Description: Block_cache size for RocksDB (block size 1024)

• Command line: --rocksdb-block-cache-size=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 536870912

• Range: 1024 to 9223372036854775807

To see the statistics of block cache usage, check SHOW ENGINE ROCKSDB STATUS output (search for lines starting with rocksdb.block.cache).

One can check the size of data of the block cache in DB_BLOCK_CACHE_USAGE column of the INFORMATION_SCHEMA.ROCKSDB_DBSTATS table.

rocksdb_block_restart_interval

• Description: BlockBasedTableOptions::block_restart_interval for RocksDB

• Command line: --rocksdb-block-restart-interval=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 16

• Range: 1 to 2147483647

rocksdb_block_size

• Description: BlockBasedTableOptions::block_size for RocksDB

• Command line: --rocksdb-block-size=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 4096

• Range: 1 to 18446744073709551615

rocksdb_block_size_deviation

• Description: BlockBasedTableOptions::block_size_deviation for RocksDB

• Command line: --rocksdb-block-size-deviation=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 10

• Range: 0 to 2147483647

rocksdb_bulk_load

• Description: Use bulk-load mode for inserts. This disables unique_checks and enables rocksdb_commit_in_the_middle.

• Command line: --rocksdb-bulk-load={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_bulk_load_allow_sk

• Description: Allow bulk loading of sk keys during bulk-load. Can be changed only when bulk load is disabled.

• Command line: --rocksdb-bulk-load_allow_sk={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_bulk_load_allow_unsorted

• Description: Allow unsorted input during bulk-load. Can be changed only when bulk load is disabled.

• Command line: --rocksdb-bulk-load_allow_unsorted={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_bulk_load_size

• Description: Maximum number of records in a batch for bulk-load mode.

• Command line: --rocksdb-bulk-load-size=#

• Scope: Global, Session

• Dynamic: Yes

• Data Type: numeric

• Default Value: 1000

• Range: 1 to 1073741824

rocksdb_bytes_per_sync

• Description: DBOptions::bytes_per_sync for RocksDB.

• Command line: --rocksdb-bytes-per-sync=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 18446744073709551615

rocksdb_cache_dump

• Description: Include RocksDB block cache content in core dump.

• Command line: --rocksdb-cache-dump={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

rocksdb_cache_high_pri_pool_ratio

• Description: Specify the size of block cache high-pri pool.

• Command line: --rocksdb-cache-high-pri-pool-ratio=#

• Scope: Global

• Dynamic: Yes

• Data Type: double

• Default Value: 0.000000

• Range: 0 to 1

rocksdb_cache_index_and_filter_blocks

• Description: BlockBasedTableOptions::cache_index_and_filter_blocks for RocksDB.

• Command line: --rocksdb-cache-index-and-filter-blocks={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_cache_index_and_filter_with_high_priority

• Description: cache_index_and_filter_blocks_with_high_priority for RocksDB.

• Command line: --rocksdb-cache-index-and-filter-with-high-priority={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_checksums_pct

• Description: Percentage of rows to be checksummed.

• Command line: --rocksdb-checksums-pct=#

• Scope: Global, Session

• Dynamic: Yes

• Data Type: numeric

• Default Value: 100

• Range: 0 to 100

rocksdb_collect_sst_properties

• Description: Enables collecting SST file properties on each flush.

• Command line: --rocksdb-collect-sst-properties={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_commit_in_the_middle

• Description: Commit rows implicitly every rocksdb_bulk_load_size, on bulk load/insert, update and delete.

• Command line: --rocksdb-commit-in-the-middle={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_commit_time_batch_for_recovery

• Description: TransactionOptions::commit_time_batch_for_recovery for RocksDB.

• Command line: --rocksdb-commit-time-batch-for-recovery={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_compact_cf

• Description: Compact column family.

• Command line: --rocksdb-compact-cf=value

• Scope: Global

• Dynamic: Yes

• Data Type: string

• Default Value: (Empty)

rocksdb_compaction_readahead_size

• Description: DBOptions::compaction_readahead_size for RocksDB.

• Command line: --rocksdb-compaction-readahead-size=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 18446744073709551615

rocksdb_compaction_sequential_deletes

• Description: RocksDB will trigger compaction for the file if it has more than this number sequential deletes per window.

• Command line: --rocksdb-compaction-sequential-deletes=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 2000000

rocksdb_compaction_sequential_deletes_count_sd

• Description: Counting SingleDelete as rocksdb_compaction_sequential_deletes.

• Command line: --rocksdb-compaction-sequential-deletes-count-sd={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_compaction_sequential_deletes_file_size

• Description: Minimum file size required for compaction_sequential_deletes.

• Command line: --rocksdb-compaction-sequential-deletes-file-size=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: -1 to 9223372036854775807

rocksdb_compaction_sequential_deletes_window

• Description: Size of the window for counting rocksdb_compaction_sequential_deletes.

• Command line: --rocksdb-compaction-sequential-deletes-window=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 2000000

rocksdb_concurrent_prepare

• Description: DBOptions::concurrent_prepare for RocksDB.

• Command line: --rocksdb-coconcurrent-prepare={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: 1

• Removed: MariaDB 10.3.7, MariaDB 10.2.15

rocksdb_create_checkpoint

• Description: Checkpoint directory.

• Command line: --rocksdb-create-checkpoint=value

• Scope: Global

• Dynamic: Yes

• Data Type: string

• Default Value: (Empty)

rocksdb_create_if_missing

• Description: DBOptions::create_if_missing for RocksDB.

• Command line: --rocksdb-create-if-missing={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_create_missing_column_families

• Description: DBOptions::create_missing_column_families for RocksDB.

• Command line: --rocksdb-create-missing-column-families={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_datadir

• Description: RocksDB data directory.

• Command line: --rocksdb-datadir[=value]

• Scope: Global

• Dynamic: No

• Data Type: string

• Default Value: ./#rocksdb

rocksdb_db_write_buffer_size

• Description: DBOptions::db_write_buffer_size for RocksDB.

• Command line: --rocksdb-db-write-buffer-size=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 0

• Range: 0 to 18446744073709551615

rocksdb_deadlock_detect

• Description: Enables deadlock detection.

• Command line: --rocksdb-deadlock-detect={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_deadlock_detect_depth

• Description: Number of transactions deadlock detection will traverse through before assuming deadlock.

• Command line: --rocksdb-deadlock-detect-depth=#

• Scope: Global, Session

• Dynamic: Yes

• Data Type: numeric

• Default Value: 50

• Range: 2 to 18446744073709551615

rocksdb_debug_manual_compaction_delay

• Description: For debugging purposes only. Sleeping specified seconds for simulating long running compactions.

• Command line: --rocksdb-debug_manual_compaction_delay=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 4294967295

rocksdb_debug_optimizer_no_zero_cardinality

• Description: If cardinality is zero, override it with some value.

• Command line: --rocksdb-debug-optimizer-no-zero-cardinality={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

rocksdb_debug_ttl_ignore_pk

• Description: For debugging purposes only. If true, compaction filtering will not occur on PK TTL data. This variable is a no-op in non-debug builds.

• Command line: --rocksdb-debug-ttl-ignore-pk={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_debug_ttl_read_filter_ts

• Description: For debugging purposes only. Overrides the TTL read filtering time to time + debug_ttl_read_filter_ts. A value of 0 denotes that the variable is not set. This variable is a no-op in non-debug builds.

• Command line: --rocksdb-debug-ttl-read-filter-ts=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: -3600 to 3600

rocksdb_debug_ttl_rec_ts

• Description: For debugging purposes only. Overrides the TTL of records to now() + debug_ttl_rec_ts. The value can be +/- to simulate a record inserted in the past vs a record inserted in the 'future'. A value of 0 denotes that the variable is not set. This variable is a no-op in non-debug builds.

• Command line: --rocksdb-debug-ttl-read-filter-ts=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: -3600 to 3600

rocksdb_debug_ttl_snapshot_ts

• Description: For debugging purposes only. Sets the snapshot during compaction to now() + debug_set_ttl_snapshot_ts. The value can be positive or negative to simulate a snapshot in the past vs a snapshot created in the 'future'. A value of 0 denotes that the variable is not set. This variable is a no-op in non-debug builds.

• Command line: --rocksdb-debug-ttl-snapshot-ts=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: -3600 to 3600

• Introduced: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_default_cf_options

• Description: Default cf options for RocksDB.

• Command line: --rocksdb-default-cf-options=value

• Scope: Global

• Dynamic: No

• Data Type: string

• Default Value: (Empty)

rocksdb_delayed_write_rate

• Description: DBOptions::delayed_write_rate.

• Command line: --rocksdb-delayed-write-rate=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0 (Previously 16777216)

• Range: 0 to 18446744073709551615

rocksdb_delete_cf

• Description: Delete column family.

• Command line: --rocksdb-delete-cf=val

• Scope: Global

• Dynamic: No

• Data Type: string

• Default Value: (Empty string)

• Introduced: MariaDB 10.4.7, MariaDB 10.3.17, MariaDB 10.2.26

rocksdb_delete_obsolete_files_period_micros

• Description: DBOptions::delete_obsolete_files_period_micros for RocksDB.

• Command line: --rocksdb-delete-obsolete-files-period-micros=#

• Scope: Global

• Dynamic: No

• Data Type: numeric

• Default Value: 21600000000

• Range: 0 to 9223372036854775807

rocksdb_enable_2pc

• Description: Enable two phase commit for MyRocks. When set, MyRocks will keep its data consistent with the binary log (in other words, the server are a crash-safe master). The consistency is achieved by doing two-phase XA commit with the binary log.

• Command line: --rocksdb-enable-2pc={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

rocksdb_enable_bulk_load_api

• Description: Enables using SstFileWriter for bulk loading.

• Command line: --rocksdb-enable-bulk-load-api={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_enable_insert_with_update_caching

• Description: Whether to enable optimization where we cache the read from a failed insertion attempt in INSERT ON DUPLICATE KEY UPDATE.

• Command line: --rocksdb-enable-insert-with-update-caching={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

• Introduced: MariaDB 10.4.7, MariaDB 10.3.17, MariaDB 10.2.26

rocksdb_enable_thread_tracking

• Description: DBOptions::enable_thread_tracking for RocksDB.

• Command line: --rocksdb-enable-thread-tracking={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_enable_ttl

• Description: Enable expired TTL records to be dropped during compaction.

• Command line: --rocksdb-enable-ttl={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

• Introduced: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_enable_ttl_read_filtering

• Description: For tables with TTL, expired records are skipped/filtered out during processing and in query results. Disabling this will allow these records to be seen, but as a result rows may disappear in the middle of transactions as they are dropped during compaction. Use with caution.

• Command line: --rocksdb-enable-ttl-read-filtering={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

• Introduced: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_enable_write_thread_adaptive_yield

• Description: DBOptions::enable_write_thread_adaptive_yield for RocksDB.

• Command line: --rocksdb-enable-write-thread-adaptive-yield={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_error_if_exists

• Description: DBOptions::error_if_exists for RocksDBB.

• Command line: --rocksdb-error-if-exists={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: OFF

rocksdb_error_on_suboptimal_collation

• Description: Raise an error instead of warning if a sub-optimal collation is used.

• Command line: --rocksdb-error-on-suboptimal-collation={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

• Introduced: MariaDB 10.3.10, MariaDB 10.2.18

rocksdb_flush_log_at_trx_commit

• Description: Sync on transaction commit. Similar to innodb_flush_log_at_trx_commit. One can check the flushing by examining the rocksdb_wal_synced and rocksdb_wal_bytes status variables.

  • 1: Always sync on commit (the default).

  • 0: Never sync.

  • 2: Sync based on a timer controlled via rocksdb-background-sync.

• Command line: --rocksdb-flush-log-at-trx-commit=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 1

• Range: 0 to 2

rocksdb_flush_memtable_on_analyze

• Description: Forces memtable flush on ANALZYE table to get accurate cardinality.

• Command line: --rocksdb-flush-memtable-on-analyze={0|1}

• Scope: Global, Session

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

• Removed: MariaDB 10.3.7, MariaDB 10.2.15

rocksdb_force_compute_memtable_stats

• Description: Force to always compute memtable stats.

• Command line: --rocksdb-force-compute-memtable-stats={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: ON

rocksdb_force_compute_memtable_stats_cachetime

• Description: Time in usecs to cache memtable estimates.

• Command line: --rocksdb-force-compute-memtable-stats-cachetime=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 60000000

• Range: 0 to 2147483647

rocksdb_force_flush_memtable_and_lzero_now

• Description: Acts similar to force_flush_memtable_now, but also compacts all L0 files.

• Command line: --rocksdb-force-flush-memtable-and-lzero-now={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

• Introduced: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_force_flush_memtable_now

• Description: Forces memstore flush which may block all write requests so be careful.

• Command line: --rocksdb-force-flush-memtable-now={0|1}

• Scope: Global

• Dynamic: Yes

• Data Type: boolean

• Default Value: OFF

rocksdb_force_index_records_in_range

• Description: Used to override the result of records_in_range() when FORCE INDEX is used.

• Command line: --rocksdb-force-index-records-in-range=#

• Scope: Global, Session

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Range: 0 to 2147483647

rocksdb_git_hash

• Description: Git revision of the RocksDB library used by MyRocks.

• Command line: --rocksdb-git-hash=value=#

• Scope: Global

• Dynamic: No

• Data Type: string

• Default Value: As per git revision.

rocksdb_hash_index_allow_collision

• Description: BlockBasedTableOptions::hash_index_allow_collision for RocksDB.

• Command line: --rocksdb-hash-index-allow-collision={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

rocksdb_ignore_unknown_options

• Description: Enable ignoring unknown options passed to RocksDB.

• Command line: --rocksdb-ignore-unknown-options={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON

• Introduced: MariaDB 10.3.7, MariaDB 10.2.15

rocksdb_index_type

• Description: BlockBasedTableOptions::index_type for RocksDB.

• Command line: --rocksdb-index-type=value

• Scope: Global

• Dynamic: No

• Data Type: enum

• Default Value: kBinarySearch

• Valid Values: kBinarySearch, kHashSearch

rocksdb_info_log_level

• Description: Filter level for info logs to be written mysqld error log. Valid values include 'debug_level', 'info_level', 'warn_level', 'error_level' and 'fatal_level'.

• Command line: --rocksdb-info-log-level=value

• Scope: Global

• Dynamic: Yes

• Data Type: enum

• Default Value: error_level

• Valid Values: error_level, debug_level, info_level, warn_level, fatal_level

rocksdb_io_write_timeout

• Description: Timeout for experimental I/O watchdog.

• Command line: --rocksdb-io-write-timeout=#

• Scope: Global

• Dynamic: Yes

• Data Type: numeric

• Default Value: 0

• Valid Values: 0 to 4294967295

• Introduced: MariaDB 10.3.1, MariaDB 10.2.8

rocksdb_is_fd_close_on_exec

• Description: DBOptions::is_fd_close_on_exec for RocksDB.

• Command line: --rocksdb-is-fd-close-on-exec={0|1}

• Scope: Global

• Dynamic: No

• Data Type: boolean

• Default Value: ON