[client]
port = 3306
socket = /var/run/mysqld/mysqld.sock
[mysqld_safe]
thp-setting=never
socket = /var/run/mysqld/mysqld.sock
nice = 0
[mysqld]
user = mysql
pid-file = /var/run/mysqld/mysqld.pid
socket = /var/run/mysqld/mysqld.sock
port = 3306
basedir = /usr
datadir = /var/lib/mysql
tmpdir = /tmp
local-infile = 0
lc-messages-dir = /usr/share/mysql
skip-external-locking
skip-name-resolve
#bind-address = 127.0.0.1
#对timestamp新的行为控制参数，相关文章。
explicit_defaults_for_timestamp = 1
character-set-server = utf8mb4
default-storage-engine=INNODB
##thread相关参数:Thread_pool_size X Thread_pool_oversubscribe = MySQL内部同时支持运行的最大任务数，相关文章
#MySQL线程调度方式,
thread_handling = pool-of-threads
#根据系统的CPU创建出thread_pool_size(该参数不建议调整)threadpool中group数量，默认为cpu核心数，server启动时自动计算
#thread_pool_size = auto
#每个thread_pool_size里同时可以运行多少个任务，超过限制后，继续创建worker线程会被延迟
thread_pool_oversubscribe = 20
#timer线程检测间隔，单位为毫秒，用于防止thread pool里的线程死掉．Thread pool里线程如果超过thread_pool_stall_limit声明的时间没响应降会重新创建
thread_pool_stall_limit = 50
#用于声明extra_port最大允许的连接，extra_port 去监听一个和服务端口不一样的端口，用于Thread pool被占满或是被锁着的情况下用于管理使用．
extra_max_connection = 5
#线程池的优先级：transactions：使用优先队列和普通队列，对于事务已经开启的statement，放到优先队列中，否则放到普通队列中。statements：只使用优先队列
；none: 只是用普通队列，本质上和statements相同，都是只是用一个队列
thread_pool_high_prio_mode = transactions
#当开启了优先队列模式后，每个连接最多允许thread_pool_high_prio_tickets次被放到优先队列中，之后放到普通队列中
thread_pool_high_prio_tickets=4294967295
#线程最大空闲时间，单位为秒，超过限制后会退出，默认60
thread_pool_idle_timeout=60
#threadpool中最大线程数目，线程总数超过该限制后不能继续创建更多线程
thread_pool_max_threads = 100000
#FullText
#MyISAM
ft_min_word_len = 1
ft_max_word_len = 84
#InnoDB
innodb_ft_min_token_size= 1
innodb_ft_max_token_size=84
#MyISAM 停止词存放在文件
ft_stopword_file = /var/lib/mysql/stopword.txt
#InnoDB，停止词存放在表
innodb_ft_server_stopword_table = dba_test/ft_stopwords
innodb_ft_user_stopword_table = dba_test/ft_stopwords
#控制在创建临时表时使用的存储引擎，默认为innodb
default_tmp_storage_engine = innodb
#线上可以开启
innodb_undo_directory = /var/lib/mysql/undolog/
innodb_undo_tablespaces = 8
innodb_undo_logs = 128
#用于优化in()，以确认是否直接使用索引统计，相关文章
eq_range_index_dive_limit = 200
#back_log = 500
thread_stack = 192K
max_allowed_packet = 32M
max_connections = 600
max_connect_errors = 2000
max_heap_table_size = 256M
max_length_for_sort_data= 2048
tmp_table_size = 256M
read_buffer_size = 512K
read_rnd_buffer_size = 1M
sort_buffer_size = 512K
join_buffer_size = 256K
table_open_cache = 1024
#对table cache进行划分，减少table cache的锁竞争，累死innodb_buffer_poll_instances
table_open_cache_instances = 8
thread_cache_size = 64
# * Query Cache Configuration
#query_cache_type = 0
#query_cache_size = 0M
#query_cache_limit = 1M
open_files_limit = 5000
# * Logging and Replication
log_error = /var/log/mysql/error_3306.log
#开启慢查询日志，相关文章1，文章2
slow_query_log = 1
slow_query_log_file = /var/log/mysql/mysql-slow.log
long_query_time = 0.5
#log-queries-not-using-indexes
#general_log = 1
#慢查询过滤器，用于Percona，相关文章
#log_slow_filter = 'string'
#可用值session or query
log_slow_rate_type = session
#记录每个session/query的日志，如果log_slow_rate_type为session，将不记录replication thread的日志
log_slow_rate_limit = 1
#语句执行存储过程(不包括触发器)慢将记录到慢查询日志
log_slow_sp_statements = 1
#指定慢查询日志包含多少信息，相关文章
#log_slow_verbosity = ''
#每个慢查询记录时间戳
#slow_query_log_timestamp_always
#精度为second或microsecond
slow_query_log_timestamp_precision=second
#该变量控制最大slow文件数量，设置slow_query_log_file为没有.log后缀的文件
#max_slowlog_files
#超过该值，slow日志将自动旋转，生成slow_query_log_file.000001这样的文件
#max_slowlog_size
log_bin = mysql-bin-3306.log
binlog_cache_size = 1M
max_binlog_size = 1024M
binlog_format = MIXED
#binlog_format = ROW
#binlog_row_image = minimal
#binlog_rows_query_log_events = 1
sync_binlog = 1000
expire_logs_days = 5#binlog_checksum = NONE
#事务提交顺序和binlog记录顺序是相同的，默认打开
binlog_order_commits =1
server-id = 10
relay_log = mysqld-relay-bin
# * TokuDB
tokudb_cache_size = 512M
tokudb_commit_sync = 0
tokudb_fsync_log_period = 1000
tokudb_data_dir = /var/lib/mysql/tokudb_data
tokudb_log_dir = /var/lib/mysql/tokudb_log
tokudb_tmp_dir = /tmp
# * MyISAM
key_buffer_size = 128M
bulk_insert_buffer_size = 8M
concurrent_insert = 2
#myisam-recover = BACKUP
myisam_recover_options = BACKUP
# * InnoDB
#打印死锁到errorlog
innodb_print_all_deadlocks = 1
#预读，对于随机读写，推荐关闭
innodb_read_ahead_threshold = 0
#预热
innodb_buffer_pool_dump_now = 1
innodb_buffer_pool_load_now = 1
innodb_buffer_pool_dump_at_shutdown = 1
innodb_buffer_pool_load_at_startup = 1
#预热文件
innodb_buffer_pool_filename = ib_buffer_pool
#事务日志记录的大小，默认512，ssd可以设置4096
innodb_log_block_size = 4096
innodb_page_size = 16K
#innodb_page_size = 4K #ssd
#innodb_flush_neighbors = 0
#ssd
innodb_flush_neighbors = 0
#影响page cleaner 线程一次扫描LRU/UNZIP_LRU的深度，默认为1024，IO能力强的可以适当调大。 innodb_io_capacity/innodb_buffer_pool_instances
innodb_lru_scan_depth =2000
innodb_strict_mode = 1
innodb_buffer_pool_size = 10000M
innodb_buffer_pool_instances = 3
innodb_change_buffering = all
innodb_change_buffer_max_size = 20
innodb_autoextend_increment = 64
innodb_additional_mem_pool_size = 15M#5.7中移除，innodb_use_sys_malloc 配置为 ON 时，innodb_additional_mem_pool_size 失效（直接从操作系统分配内存）
innodb_file_per_table = 1
innodb_flush_log_at_trx_commit = 2
#自定义刷新日志时间，每隔这么多秒刷一次日志，只有在innodb_flush_log_at_trx_commit=2时才生效
innodb_flush_log_at_timeout = 3
innodb_log_buffer_size = 8M
innodb_log_file_size = 256M
innodb_log_files_in_group = 2
innodb_max_dirty_pages_pct = 60
innodb_read_io_threads = 8
innodb_write_io_threads = 8
#innodb_io_capacity = 800
innodb_io_capacity = 2000 #ssd
#innodb_io_capacity_max = 2000
innodb_adaptive_flushing = 1
innodb_adaptive_flushing_lwm = 15
innodb_old_blocks_pct = 20
innodb_old_blocks_time = 1000
#online ddl时并发DML产生的row log最大size，超过这个限制会导致DDL回滚
innodb_online_alter_log_max_size = 128M
innodb_flush_neighbors = 1
innodb_concurrency_tickets = 5000
innodb_optimize_fulltext_only = 0
innodb_use_native_aio = 1
innodb_use_sys_malloc = 1#innodb_use_sys_malloc 配置为 ON 时，innodb_additional_mem_pool_size 失效（直接从操作系统分配内存）。5.7中移除
innodb_thread_concurrency = 0#回收undo线程
innodb_purge_threads= 4
#5.7
#innodb_page_cleaners = 4
#每次回收undo页的数量
innodb_purge_batch_size = 300
#创建索引做merge-sort时，一个归并block的大小
innodb_sort_buffer_size = 16M
innodb_change_buffering = all
#changebuffer占用bp的最大比例
innodb_change_buffer_max_size = 20
innodb_file_format = barracuda
innodb_file_format_check = 1
innodb_file_format_max = barracuda
innodb_stats_on_metadata = 0
innodb_open_files = 3000
#Percona支持ALL_O_DIRECT
innodb_flush_method = ALL_O_DIRECT
#innodb_flush_method = O_DIRECT_NO_FSYNC
lock_wait_timeout = 50
transaction-isolation = READ-COMMITTED
# * replication
replicate-ignore-db = test
replicate-ignore-db = mysql
#binlog_do_db = include_database_name
skip-slave-start = 1
log_slave_updates = 1
#gtid-mode=on # GTID only
#enforce-gtid-consistency=true # GTID only
master_info_repository = table
relay_log_info_repository = table
# * slave
#并行复制线程数
slave_parallel_workers = 4
#多线程复制需要关闭
slave_transaction_retries = 0
#在并发复制时总共执行这么多次事务后做一次checkpoint，更新show slave status的数据
slave_checkpoint_group = 512
#在复制执行这么长时间后做一次checkpoint
slave_checkpoint_period = 300
#在多线程复制时，在队列中Pending的事件所占用的最大内存，默认为16M，如果内存富余，或者延迟较大时，可以适当调大;注意这个值要比主库的max_allowed_packet大
slave_pending_jobs_size_max = 32M
#备库SQL线程是否检查binlog的checksum
slave_sql_verify_checksum = 1
#部分解决无主键表导致的复制延迟问题，相关文章
#slave_rows_search_algorithms = TABLE_SCAN,INDEX_SCAN,HASH_SCAN
#slave_net_timeout = 300
relay_log_recovery = 1
#sync_relay_log = 0
#sync_relay_log_info = 0
#sync_master_info = 0
# ssl-ca=/etc/mysql/cacert.pem
# ssl-cert=/etc/mysql/server-cert.pem
# ssl-key=/etc/mysql/server-key.pem
[mysqldump]
quick
quote-names
max_allowed_packet = 16M
[mysql]
#no-auto-rehash # faster start of mysql but no tab completition
[isamchk]
key_buffer = 16M
#
# * IMPORTANT: Additional settings that can override those from this file!
# The files must end with '.cnf', otherwise they'll be ignored.
#
!includedir /etc/mysql/conf.d/
该文章后续会不断更新，慢慢完善。之前也介绍了MySQL5.5的配置文件，现在按照前2篇文章1和文章2的信息，整理出该博文。对MySQL5.6的my.cnf 进行了配置(Percona MySQL)，因为根据服务器硬件资源来调整的，所以一些内存、io刷写的大小比例可以根据自己的条件来设置。
相关文章：
http://dev.mysql.com/doc/refman/5.6/en/innodb-parameters.html
https://www.percona.com/doc/percona-server/5.6/ps-versions-comparison.html
https://dev.mysql.com/doc/refman/5.6/en/server-system-variables.html