Skip to content

Latest commit

 

History

History
398 lines (344 loc) · 13.5 KB

INTERNALS.md

File metadata and controls

398 lines (344 loc) · 13.5 KB

MySQLTuner Internals

Table of contents

MySQLTuner steps

  • Header Print
  • Get login information
  • Set up some OS variables
  • Toss variables/status into hashes
  • Get information about the tuning connexion
  • Check current MySQL version
  • Suggest 64-bit upgrade
  • Show enabled storage engines
  • Show informations about databases (option: --dbstat)
  • Show informations about indexes (option: --idxstat)
  • Display some security recommendations
  • Calculate everything we need
  • Print the server stats
  • Print MyISAM stats
  • Print InnoDB stats
  • Print AriaDB stats
  • Print replication info
  • Make recommendations based on stats
  • Close reportfile if needed
  • Dump result if debug is on

MySQLTuner get login information steps

  • Is a login possible?
    • Force socket?
    • Remote connection?
      • Specifying available RAM is required
    • Got user/pass on command line?
    • mysql-quickbackup credentials available?
    • Plesk credentials available?
    • DirectAdmin credentials available?
    • Debian maintenance account credentials available?
    • Just try a login
      • If working, and .my.cnf isn't there, WARNING
      • If working, and .my.cnf is there, okay
    • Prompt for creds on the console

MySQLTuner system checks

  • 32-bit w/>2GB RAM check
  • Check number of opened ports (warning if more than 9 ports opened)
  • Check 80, 8080, 443 and 8443 ports if warning are raised if there are opened
  • Check if some banned ports are not opened (option --bannedports separated by comma)
  • Check if non kernel and user process except mysqld are not using more than 15% of total physical memory)
  • Check vm.swapiness
  • Check /etc/security/limit.conf
  • Check sysctl entries: sunrpc.tcp_slot_entries, vm.swappiness, fs.aio-fs-nr
  • Check mount point
  • Check Ethernet card
  • Check load average

MySQLTuner Server version checks

  • EOL MySQL version check
  • Currently MySQL < 5.1 are EOF considerated.
  • Using 5.5+ version of MySQL for performance issue (asynchronous IO).

MySQL Storage engine general information

  • Get storage engine counts/stats
    • Check for DB engines that are enabled but unused
    • Look for fragmented tables
    • Look for auto-increments near capacity
      • Look for table with autoincrement with value near max capacity

MySQLTuner security checks

  • Is anonymous user present?
  • Users without passwords
  • Users w/username as password
  • Users w/o host restriction
  • Weak password check (possibly using cracklib later?)
  • Using basic_passwords.txt as password database
  • Password list checks can be avoid (option: --skippassword)

MySQLTuner CVE vulnerabilities detection

  • option: --cvefile
  • Check if your MariaDB or MySQL version contains CVE bulletins.

MySQLTuner database information

  • Per database information * Tables number
    • Rows number
    • Total size
    • Data size
    • Percentage of data size
    • Index size
    • Percentage of index size
      • Collation number
      • Check that there is only one collation for all table in a database
      • Check that there is only one collation for ll table columns in a database
      • Check that there is only one storage engine per user database

MySQLTuner index information

  • Top 10 worth selectivity index
  • Per index information
    • Index Cardinality
    • Index Selectivity
    • Misc information about index definition
    • Misc information about index size

MySQLTuner Connections information

  • Highest usage of available connections
  • Percentage of used connections (<85%)
  • Percentage of aborted connections (<3%)

MySQLTuner server information

  • Uptime: If MySQL started within last 24 hours
  • Bytes received and sent
  • Number of connections
  • Percentage between reads and writes
  • Is binary log activated ?
    • Is GTID mode activated ?

MySQLTuner sort, join and temp table information

  • Max memory temporary table size allowed.
  • Percentage of sort using temporary table (<10%)
  • Number of join performed without using indexes (<250)
  • Percentage of temporary table written on disk(<25%)
  • Thread cache (=4)
  • Thread cache hit ratio (>50%) if thread_handling is different of pools-of-threads
  • Table cache hit ratio(>2°%)
  • Percentage of open file and open file limit(<85%)
  • Percentage of table locks (<95%)
  • Percentage of binlog cache lock (<90%)

MySQLTuner global buffer information

  • Key Buffer
  • Max Tmp Table
  • Per Thread Buffer
    • Read Buffer
    • Read RND Buffer
    • Sort Buffer
    • Thread stack
    • Join Buffer
    • Binlog Cache Buffers size if activated

MySQLTuner query cache checks

  • Is Query cache activated ?
    • Query Cache Buffers
    • Query Cache DISABLED, ALL REQUEST or ON DEMAND
    • Query Cache Size
    • Query cache hit ratio (cache efficienty)

MySQLTuner memory checks

  • Get total RAM/swap
  • Is there enough memory for max connections reached by MySQL ?
  • Is there enough memory for max connections allowed by MySQL ?
  • Max percentage of memory used(<85%)

MySQLTuner slow queries checks

  • Percentage of Slow queries (<5%)

MySQLTuner replication checks

  • Is server replication configuarted as slave ?
  • SQL replacation thread running ?
  • IO replacation thread running ?
  • Replication lag in seconds
  • Is Slave configuratedd in read only ?

MySQLTuner InnoDB information

  • InnoDB Buffer Pool Size
    • If possible, innodb_buffer_pool_size should be greater data and index size for Innodb Table
    • Innodb_buffer_pool_size should around 75 to 80 % of the available system memory.
  • InnoDB Buffer Pool Instances
    • MySQL needs 1 instanes per 1Go of Buffer Pool
    • innodb_buffer_pool instances = round(innodb_buffer_pool_size / 1Go)
    • innodb_buffer_pool instances must be equals or lower than 64
  • InnoDB Buffer Pool Usage
    • If more than 20% of InnoDB buffer pool is not used, MySQLTuner raise an alert.
  • InnoDB Buffer Pool Log Size
    • InnoDB total log file size should be 25% of innodb_buffer_pool_size
  • InnoDB Read effiency
    • Ratio of read without locks
  • InnoDB Write effiency
    • Ratio of write without locks
  • InnoDB Log Waits
    • Checks that no lock is used on Innodb Log.
  • InnoDB Chunk Size
    • Check InnoDB Buffer Pool size is a multiple of InnoDB Buffer Pool chunk size * InnoDB Buffer Pool instances

MySQLTuner ARIADB information

  • Is Aria indexes size is greater than page cache size ?
  • Page cache read hit ratio (>95%)
  • Page cache write hit ratio (>95%)

MySQLTuner MYISAM information

  • Key buffer usage (>90%)
  • Is MyISAM indexes size is greater than key buffer size ?
  • Key buffer read hit ratio (>95%)
  • Key buffer write hit ratio (>95%)

MySQLTuner Galera information

  • wsrep_ready cluster is ready
  • wsrep_connected node is connected to other nodes
  • wsrep_cluster_name is defined.
  • wsrep_node_name is defined.
  • Check thet notification script wsrep_notify_cmd is defined
  • wsrep_cluster_status PRIMARY /NON PRIMARY.
    • PRIMARY : Coherent cluster
    • NO PRIMARY : cluster gets several states
  • wsrep_ local_state_comment: Node state
    • SYNCED (uptodate),
    • DONOR(sending information to another node)
    • Joiner(Try to reach cluster group)
    • SYNCED state able to read/write
  • wsrep_cluster_conf_id configuration level must be identical in all nodes
  • wsrep_slave_thread is between 3 or 4 times number of CPU core.
  • gcs.limit should be equal to wsrep_slave_threads * 5
  • gcs.fc_factor should be equal to 0.8
  • Flow control fraction should be lower than 0.02 (wsrep_flow_control_paused < 0.02)
  • wsrep_last_commited committed level must be identical in all nodes
  • Look for tables without primary keys
  • Look for non InnoDB tables for Galera
  • Variable innodb_flush_log_at_trx_commit should be set to 0.
  • Check that there is 3 or 5 members in Galera cluster.
  • Check that xtrabackup is used for SST method with wsrep_sst_method variable.
  • Check variables wsrep_OSU_method is defined to TOI for updates.
  • Check that there is no certification failures controlling wsrep_local_cert_failures status.

MySQLTuner TokuDB information

  • tokudb_cache_size
  • tokudb_directio
  • tokudb_empty_scan
  • tokudb_read_block_size
  • tokudb_commit_sync
  • tokudb_checkpointing_period
  • tokudb_block_size
  • tokudb_cleaner_iterations
  • tokudb_fanout

MySQLTuner XtraDB information

  • Nothing for the moment

MySQLTuner Connect information

  • Nothing for the moment

MySQLTuner Spider information

  • Nothing for the moment

MySQLTuner RocksDb information

  • Nothing for the moment

MySQLTuner Thread pool information

  • thread_pool_size between 16 to 36 for Innodb usage
  • thread_pool_size between 4 to 8 for MyIsam usage

MySQLTuner performance schema and sysschema information

  • Check that Performance schema is activated for 5.6+ version
  • Check that Performance schema is disactivated for 5.5- version
  • Check that Sys schema is installed
  • sys Schema version
  • Top user per connection
  • Top user per statement
  • Top user per statement latency
  • Top user per lock latency
  • Top user per full scans
  • Top user per row_sent
  • Top user per row modified
  • Top user per io
  • Top user per io latency
  • Top host per connection
  • Top host per statement
  • Top host per statement latency
  • Top host per lock latency
  • Top host per full scans
  • Top host per rows sent
  • Top host per rows modified
  • Top host per io
  • Top 5 host per io latency
  • Top IO type order by total io
  • Top IO type order by total latency
  • Top IO type order by max latency
  • Top Stages order by total io
  • Top Stages order by total latency
  • Top Stages order by avg latency
  • Top host per table scans
  • InnoDB Buffer Pool by schema
  • InnoDB Buffer Pool by table
  • Process per allocated memory
  • InnoDB Lock Waits
  • Threads IO Latency
  • High Cost SQL statements
  • Top 5% slower queries
  • Top 10 nb statement type
  • Top statement by total latency
  • Top statement by lock latency
  • Top statement by full scans
  • Top statement by rows sent
  • Top statement by rows modified
  • Use temporary tables
  • Unused Indexes
  • Full table scans
  • Latest file IO by latency
  • File by IO read bytes
  • File by IO written bytes
  • File per IO total latency
  • File per IO read latency
  • File per IO write latency
  • Event Wait by read bytes
  • Event Wait by write bytes
  • Event per wait total latency
  • Event per wait read latency
  • Event per wait write latency
  • TOP 15 most read indexes
  • TOP 15 most modified indexes
  • TOP 15 high select latency index
  • TOP 15 high insert latency index
  • TOP 15 high update latency index
  • TOP 15 high delete latency index
  • TOP 15 most read tables
  • TOP 15 most modified tables
  • TOP 15 high select latency tables
  • TOP 15 high insert latency tables
  • TOP 15 high update latency tables
  • TOP 15 high delete latency tables
  • Redundant indexes
  • Tables not using InnoDb buffer
  • Top 15 Tables using InnoDb buffer
  • Top 15 Tables with InnoDb buffer free
  • Top 15 Most executed queries
  • Latest SQL queries in errors or warnings
  • Top 20 queries with full table scans
  • TOP 15 reader queries (95% percentile)
  • TOP 15 row look queries (95% percentile)
  • TOP 15 total latency queries (95% percentile)
  • TOP 15 max latency queries (95% percentile)
  • TOP 15 average latency queries (95% percentile)
  • Top 20 queries with sort
  • Last 50 queries with sort
  • TOP 15 row sorting queries with sort
  • TOP 15 total latency queries with sort
  • TOP 15 merge queries with sort
  • TOP 15 average sort merges queries with sort
  • TOP 15 scans queries with sort
  • TOP 15 range queries with sort
  • Top 20 queries with temp table
  • Last 50 queries with temp table
  • TOP 15 total latency queries with temp table
  • TOP 15 queries with temp table to disk
  • TOP 15 class events by number
  • TOP 30 events by number
  • TOP 15 class events by total latency
  • TOP 30 events by total latency
  • TOP 15 class events by max latency
  • TOP 30 events by max latency