MySQL集群PXC的搭建
最近公司某客户要求我们的数据库搭建PXC集群以保证他们的系统高性能和搞稳定性
以后花费了一些时间去搭建和测试，也踩过一些坑，准备分享出来
系统：centos6.6
PXC：5.6.26
建议关闭iptables， 4个端口 3306,4444,4567,4568
pxc环境所涉及的端口：
#mysql实例端口：3306.
#pxc cluster相互通讯的端口：4567
Port for group communication, default 4567. It can be changed by the option:
wsrep_provider_options ="gmcast.listen_addr=tcp://0.0.0.0:4010; "
#用于SST传送的端口：4444
Port for State Transfer, default 4444. It can be changed by the option:
wsrep_sst_receive_address=10.11.12.205:5555
#用于IST传送的端口：4568
Port for Incremental State Transfer, default port for group communication + 1 (4568). It can be changed by the option:
wsrep_provider_options = "ist.recv_addr=10.11.12.206:7777; "
node1 192.168.3.130
node2 192.168.3.129
node3 192.168.3.128
1.安装软件依赖包（添加repl源，再安装依赖，3台server操作一样。）
rpm -ivh http://dl.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm yum install perl-IO-Socket-SSL perl-DBD-MySQL perl-Time-HiRes socat nc -y
2.安装xtrabackup（PXC同步数据需要用到）
yum install -y http://www.percona.com/downloads/percona-release/redhat/0.1-3/percona-release-0.1-3.noarch.rpm
3.下载安装Percona-XtraDB-Cluster
#安装开发包
yum install -y libaio*
yum groupinstall -y 'Development tools'
/usr/lib64/libssl.so.
/usr/lib64/libcrypto.so.
wget https://www.percona.com/downloads/Percona-XtraDB-Cluster-56/Percona-XtraDB-Cluster-5.6.26-25.12/binary/tarball/Percona-XtraDB-Cluster-5.6.26-rel74.0-25.12.1.Linux.x86_64.tar.gz
-rel74.-.Linux.x86_64.tar.gz
useradd -M -s /sbin/nologin mysql
mkdir -p /usr/local/pxc
mkdir -p /data/pxc/mysql3306/{data,tmp,logs}
-rel74.-.Linux.x86_64/* /usr/local/pxc/
cp /usr/local/mysql/support-files/mysql.server /etc/init.d/mysql
echo "export PATH=$PATH:/usr/local/mysql/bin" >> /etc/profile
source /etc/profile
chown -R mysql:mysql /data/pxc/mysql3306/
chown -R mysql:mysql /usr/local/pxc
修改my.cnf配置文件，注意在[mysqld]段落添加PXC的参数：
vi /etc/my.cnf
130
[client] port = socket = /data/mysql/mysql3306/tmp/mysql.sock # The MySQL server [mysqld] #########Basic################## explicit_defaults_for_timestamp=true port = user = mysql basedir = /usr/local/mysql datadir = /data/mysql/mysql3306/data tmpdir = /data/mysql/mysql3306/tmp pid-file = /data/mysql/mysql3306/tmp/mysql.pid socket = /data/mysql/mysql3306/tmp/mysql.sock #skip-grant-tables #character set character_set_server = utf8 open_files_limit = back_log = #event_scheduler = ON #lower_case_table_names= skip-external-locking skip_name_resolve = default-storage-engine = InnoDB #timeout wait_timeout= interactive_timeout= connect_timeout = server- #ip最后一位+端口号= #percona 的--recursion-method slavehost模式 #report_host = 10.105.9.115 #report_port = #plugin plugin-load="semisync_master.so;semisync_slave.so" #########error log############# log-error = /data/mysql/mysql3306/logs/error.log log-warnings = #########general log############# #general_log= #general_log_file=/data/mysql/mysql3306/logs/mysql.log #########slow log############# slow_query_log = long_query_time= slow_query_log_file = /data/mysql/mysql3306/logs/mysql.slow ############# for replication################### log-bin = /data/mysql/mysql3306/logs/mysql-bin binlog_format = row max_binlog_size = 500M binlog_cache_size = 2M max_binlog_cache_size = 2M expire-logs-days = slave-net-timeout= log_bin_trust_function_creators = log-slave-updates = skip-slave-start = #read_only = #从上设置,.7是super_read_only #GTID gtid-mode = on binlog_gtid_simple_recovery= enforce_gtid_consistency= #relay log relay-log = /data/mysql/mysql3306/logs/mysql-relay relay-log-index=/data/mysql/mysql3306/logs/relay-bin.index max-relay-log-size = 500M #replication crash safe sync_master_info = sync_relay_log_info = sync_relay_log = relay_log_recovery = master_info_repository = TABLE relay_log_info_repository = TABLE #semisync 动态开启 主从切换的时候用 #rpl_semi_sync_master_enabled = #rpl_semi_sync_master_wait_no_slave = #rpl_semi_sync_master_timeout = #rpl_semi_sync_slave_enabled = #rpl_semi_sync_master_timeout = #不要超时,防止切异步,保证主从数据的完全一致性,默认是10000 #rpl_semi_sync_master_wait_point = .7新增参数 after_sync .7增强半同步 after_commit5.6普通半同步 #rpl_semi_sync_master_wait_for_slave_count = #.7新增参数 等待多少个从库接收到binlog #ignore #replicate-ignore-db = 'school','school2' #replicate-do-db = 'school','school2' #Multi-threaded Slave #slave_parallel_workers= #slave-parallel-type=LOGICAL_CLOCK #slave-parallel-type=DATABASE #replication error #slave-skip-errors=,, #######per_thread_buffers##################### max_connections= max_user_connections= max_connect_errors= #myisam_recover key_buffer_size = 64M max_allowed_packet = 16M #table_cache = table_open_cache = table_definition_cache = read_buffer_size = 1M join_buffer_size = 128K read_rnd_buffer_size = 1M #myisam sort_buffer_size = 128K myisam_max_sort_file_size = 10G myisam_repair_threads = myisam_sort_buffer_size = 32M tmp_table_size = 32M max_heap_table_size = 64M query_cache_type= query_cache_size = bulk_insert_buffer_size = 32M thread_cache_size = #thread_concurrency = thread_stack = 192K ###############InnoDB########################### innodb_data_home_dir = /data/mysql/mysql3306/data innodb_log_group_home_dir = /data/mysql/mysql3306/logs innodb_data_file_path = ibdata1:1000M:autoextend innodb_buffer_pool_size = 1G #根据内存大小设置 innodb_buffer_pool_instances = #innodb_additional_mem_pool_size = 16M innodb_log_file_size = 500M innodb_log_buffer_size = 16M innodb_log_files_in_group = innodb_flush_log_at_trx_commit = sync_binlog = innodb_lock_wait_timeout = innodb_sync_spin_loops = innodb_max_dirty_pages_pct = innodb_support_xa = innodb_thread_concurrency = innodb_thread_sleep_delay = innodb_concurrency_tickets = innodb_flush_method = O_DIRECT innodb_file_per_table = innodb_read_io_threads = innodb_write_io_threads = innodb_io_capacity = #机械盘800 ssd innodb_flush_neighbors = innodb_file_format = Barracuda innodb_purge_threads= #.6只能有一个，.7可以设置多个 innodb_purge_batch_size = innodb_old_blocks_pct= innodb_change_buffering=all innodb_stats_on_metadata=OFF innodb_print_all_deadlocks = #innodb_status_output= #innodb_status_output_locks= performance_schema= #是否打开P_S库进行性能收集 transaction_isolation = READ-COMMITTED #innodb_force_recovery= #innodb_fast_shutdown= #innodb_status_file = #实时输出show engine innodb status到innodb_status文件 ############# PXC ##################### innodb_autoinc_lock_mode= #自增锁的优化 wsrep_cluster_name=pxc-dongzheng #集群名字 wsrep_provider=/usr/local/mysql/lib/libgalera_smm.so #库文件位置 wsrep_cluster_address=gcomm://192.168.3.130,192.168.3.128,192.168.3.129 #节点中所有ip wsrep_node_address=192.168.3.130 #本节点的ip wsrep_slave_threads= # 开启的复制线程数，建议cpu核数* ，解决apply_cb跟不上问题 wsrep_sst_auth=sst:dongzheng #sst模式需要的用户名和密码 wsrep_sst_method=xtrabackup-v2 #采用什么方式复制数据。还支持mysqldump，rsync #wsrep_provider_options="debug=1;gcache.size=2G;pc.ignore_quorum=true;pc.ignore_sb=true" #打开调试模式 wsrep_provider_options="debug=1;gcache.size=2G;" wsrep_max_ws_rows= wsrep_max_ws_size= [mysqldump] quick max_allowed_packet = 128M [mysql] no-auto-rehash max_allowed_packet = 128M prompt = '(product)\u@\h:\p [\d]> ' default_character_set = utf8 [myisamchk] key_buffer_size = 64M sort_buffer_size = 512k read_buffer = 2M write_buffer = 2M [mysqlhotcopy] interactive-timeout [mysqld_safe] #malloc-lib= /usr/local/mysql/lib/mysql/libjemalloc.so
129
[client] port = socket = /data/mysql/mysql3306/tmp/mysql.sock # The MySQL server [mysqld] #########Basic################## explicit_defaults_for_timestamp=true port = user = mysql basedir = /usr/local/mysql datadir = /data/mysql/mysql3306/data tmpdir = /data/mysql/mysql3306/tmp pid-file = /data/mysql/mysql3306/tmp/mysql.pid socket = /data/mysql/mysql3306/tmp/mysql.sock #skip-grant-tables #character set character_set_server = utf8 open_files_limit = back_log = #event_scheduler = ON #lower_case_table_names= skip-external-locking skip_name_resolve = default-storage-engine = InnoDB #timeout wait_timeout= interactive_timeout= connect_timeout = server- #ip最后一位+端口号= #percona 的--recursion-method slavehost模式 #report_host = 10.105.9.115 #report_port = #plugin plugin-load="semisync_master.so;semisync_slave.so" #########error log############# log-error = /data/mysql/mysql3306/logs/error.log log-warnings = #########general log############# #general_log= #general_log_file=/data/mysql/mysql3306/logs/mysql.log #########slow log############# slow_query_log = long_query_time= slow_query_log_file = /data/mysql/mysql3306/logs/mysql.slow ############# for replication################### log-bin = /data/mysql/mysql3306/logs/mysql-bin binlog_format = row max_binlog_size = 500M binlog_cache_size = 2M max_binlog_cache_size = 2M expire-logs-days = slave-net-timeout= log_bin_trust_function_creators = log-slave-updates = skip-slave-start = #read_only = #从上设置,.7是super_read_only #GTID gtid-mode = on binlog_gtid_simple_recovery= enforce_gtid_consistency= #relay log relay-log = /data/mysql/mysql3306/logs/mysql-relay relay-log-index=/data/mysql/mysql3306/logs/relay-bin.index max-relay-log-size = 500M #replication crash safe sync_master_info = sync_relay_log_info = sync_relay_log = relay_log_recovery = master_info_repository = TABLE relay_log_info_repository = TABLE #semisync 动态开启 主从切换的时候用 #rpl_semi_sync_master_enabled = #rpl_semi_sync_master_wait_no_slave = #rpl_semi_sync_master_timeout = #rpl_semi_sync_slave_enabled = #rpl_semi_sync_master_timeout = #不要超时,防止切异步,保证主从数据的完全一致性,默认是10000 #rpl_semi_sync_master_wait_point = .7新增参数 after_sync .7增强半同步 after_commit5.6普通半同步 #rpl_semi_sync_master_wait_for_slave_count = #.7新增参数 等待多少个从库接收到binlog #ignore #replicate-ignore-db = 'school','school2' #replicate-do-db = 'school','school2' #Multi-threaded Slave #slave_parallel_workers= #slave-parallel-type=LOGICAL_CLOCK #slave-parallel-type=DATABASE #replication error #slave-skip-errors=,, #######per_thread_buffers##################### max_connections= max_user_connections= max_connect_errors= #myisam_recover key_buffer_size = 64M max_allowed_packet = 16M #table_cache = table_open_cache = table_definition_cache = read_buffer_size = 1M join_buffer_size = 128K read_rnd_buffer_size = 1M #myisam sort_buffer_size = 128K myisam_max_sort_file_size = 10G myisam_repair_threads = myisam_sort_buffer_size = 32M tmp_table_size = 32M max_heap_table_size = 64M query_cache_type= query_cache_size = bulk_insert_buffer_size = 32M thread_cache_size = #thread_concurrency = thread_stack = 192K ###############InnoDB########################### innodb_data_home_dir = /data/mysql/mysql3306/data innodb_log_group_home_dir = /data/mysql/mysql3306/logs innodb_data_file_path = ibdata1:1000M:autoextend innodb_buffer_pool_size = 1G #根据内存大小设置 innodb_buffer_pool_instances = #innodb_additional_mem_pool_size = 16M innodb_log_file_size = 500M innodb_log_buffer_size = 16M innodb_log_files_in_group = innodb_flush_log_at_trx_commit = sync_binlog = innodb_lock_wait_timeout = innodb_sync_spin_loops = innodb_max_dirty_pages_pct = innodb_support_xa = innodb_thread_concurrency = innodb_thread_sleep_delay = innodb_concurrency_tickets = innodb_flush_method = O_DIRECT innodb_file_per_table = innodb_read_io_threads = innodb_write_io_threads = innodb_io_capacity = #机械盘800 ssd innodb_flush_neighbors = innodb_file_format = Barracuda innodb_purge_threads= #.6只能有一个，.7可以设置多个 innodb_purge_batch_size = innodb_old_blocks_pct= innodb_change_buffering=all innodb_stats_on_metadata=OFF innodb_print_all_deadlocks = #innodb_status_output= #innodb_status_output_locks= performance_schema= #是否打开P_S库进行性能收集 transaction_isolation = READ-COMMITTED #innodb_force_recovery= #innodb_fast_shutdown= #innodb_status_file = #实时输出show engine innodb status到innodb_status文件 ############# PXC ##################### innodb_autoinc_lock_mode= wsrep_cluster_name=pxc-dongzheng wsrep_provider=/usr/local/mysql/lib/libgalera_smm.so wsrep_cluster_address=gcomm://192.168.3.130,192.168.3.128,192.168.3.129 wsrep_node_address=192.168.3.129 wsrep_slave_threads= wsrep_sst_auth=sst:dongzheng wsrep_sst_method=xtrabackup-v2 #wsrep_provider_options="debug=1;gcache.size=2G;pc.ignore_quorum=true;pc.ignore_sb=true" #打开调试模式 wsrep_provider_options="debug=1;gcache.size=2G;" wsrep_max_ws_rows= wsrep_max_ws_size= [mysqldump] quick max_allowed_packet = 128M [mysql] no-auto-rehash max_allowed_packet = 128M prompt = '(product)\u@\h:\p [\d]> ' default_character_set = utf8 [myisamchk] key_buffer_size = 64M sort_buffer_size = 512k read_buffer = 2M write_buffer = 2M [mysqlhotcopy] interactive-timeout [mysqld_safe] #malloc-lib= /usr/local/mysql/lib/mysql/libjemalloc.so
128
[client] port = socket = /data/mysql/mysql3306/tmp/mysql.sock # The MySQL server [mysqld] #########Basic################## explicit_defaults_for_timestamp=true port = user = mysql basedir = /usr/local/mysql datadir = /data/mysql/mysql3306/data tmpdir = /data/mysql/mysql3306/tmp pid-file = /data/mysql/mysql3306/tmp/mysql.pid socket = /data/mysql/mysql3306/tmp/mysql.sock #skip-grant-tables #character set character_set_server = utf8 open_files_limit = back_log = #event_scheduler = ON #lower_case_table_names= skip-external-locking skip_name_resolve = default-storage-engine = InnoDB #timeout wait_timeout= interactive_timeout= connect_timeout = server- #ip最后一位+端口号= #percona 的--recursion-method slavehost模式 #report_host = 10.105.9.115 #report_port = #plugin plugin-load="semisync_master.so;semisync_slave.so" #########error log############# log-error = /data/mysql/mysql3306/logs/error.log log-warnings = #########general log############# #general_log= #general_log_file=/data/mysql/mysql3306/logs/mysql.log #########slow log############# slow_query_log = long_query_time= slow_query_log_file = /data/mysql/mysql3306/logs/mysql.slow ############# for replication################### log-bin = /data/mysql/mysql3306/logs/mysql-bin binlog_format = row max_binlog_size = 500M binlog_cache_size = 2M max_binlog_cache_size = 2M expire-logs-days = slave-net-timeout= log_bin_trust_function_creators = log-slave-updates = skip-slave-start = #read_only = #从上设置,.7是super_read_only #GTID gtid-mode = on binlog_gtid_simple_recovery= enforce_gtid_consistency= #relay log relay-log = /data/mysql/mysql3306/logs/mysql-relay relay-log-index=/data/mysql/mysql3306/logs/relay-bin.index max-relay-log-size = 500M #replication crash safe sync_master_info = sync_relay_log_info = sync_relay_log = relay_log_recovery = master_info_repository = TABLE relay_log_info_repository = TABLE #semisync 动态开启 主从切换的时候用 #rpl_semi_sync_master_enabled = #rpl_semi_sync_master_wait_no_slave = #rpl_semi_sync_master_timeout = #rpl_semi_sync_slave_enabled = #rpl_semi_sync_master_timeout = #不要超时,防止切异步,保证主从数据的完全一致性,默认是10000 #rpl_semi_sync_master_wait_point = .7新增参数 after_sync .7增强半同步 after_commit5.6普通半同步 #rpl_semi_sync_master_wait_for_slave_count = #.7新增参数 等待多少个从库接收到binlog #ignore #replicate-ignore-db = 'school','school2' #replicate-do-db = 'school','school2' #Multi-threaded Slave #slave_parallel_workers= #slave-parallel-type=LOGICAL_CLOCK #slave-parallel-type=DATABASE #replication error #slave-skip-errors=,, #######per_thread_buffers##################### max_connections= max_user_connections= max_connect_errors= #myisam_recover key_buffer_size = 64M max_allowed_packet = 16M #table_cache = table_open_cache = table_definition_cache = read_buffer_size = 1M join_buffer_size = 128K read_rnd_buffer_size = 1M #myisam sort_buffer_size = 128K myisam_max_sort_file_size = 10G myisam_repair_threads = myisam_sort_buffer_size = 32M tmp_table_size = 32M max_heap_table_size = 64M query_cache_type= query_cache_size = bulk_insert_buffer_size = 32M thread_cache_size = #thread_concurrency = thread_stack = 192K ###############InnoDB########################### innodb_data_home_dir = /data/mysql/mysql3306/data innodb_log_group_home_dir = /data/mysql/mysql3306/logs innodb_data_file_path = ibdata1:1000M:autoextend innodb_buffer_pool_size = 1G #根据内存大小设置 innodb_buffer_pool_instances = #innodb_additional_mem_pool_size = 16M innodb_log_file_size = 500M innodb_log_buffer_size = 16M innodb_log_files_in_group = innodb_flush_log_at_trx_commit = sync_binlog = innodb_lock_wait_timeout = innodb_sync_spin_loops = innodb_max_dirty_pages_pct = innodb_support_xa = innodb_thread_concurrency = innodb_thread_sleep_delay = innodb_concurrency_tickets = innodb_flush_method = O_DIRECT innodb_file_per_table = innodb_read_io_threads = innodb_write_io_threads = innodb_io_capacity = #机械盘800 ssd innodb_flush_neighbors = innodb_file_format = Barracuda innodb_purge_threads= #.6只能有一个，.7可以设置多个 innodb_purge_batch_size = innodb_old_blocks_pct= innodb_change_buffering=all innodb_stats_on_metadata=OFF innodb_print_all_deadlocks = #innodb_status_output= #innodb_status_output_locks= performance_schema= #是否打开P_S库进行性能收集 transaction_isolation = READ-COMMITTED #innodb_force_recovery= #innodb_fast_shutdown= #innodb_status_file = #实时输出show engine innodb status到innodb_status文件 ############# PXC ##################### innodb_autoinc_lock_mode= wsrep_cluster_name=pxc-dongzheng wsrep_provider=/usr/local/mysql/lib/libgalera_smm.so wsrep_cluster_address=gcomm://192.168.3.130,192.168.3.128,192.168.3.129 wsrep_node_address=192.168.3.128 wsrep_slave_threads= wsrep_sst_auth=sst:dongzheng wsrep_sst_method=xtrabackup-v2 #wsrep_provider_options="debug=1;gcache.size=2G;pc.ignore_quorum=true;pc.ignore_sb=true" #打开调试模式 wsrep_provider_options="debug=1;gcache.size=2G;" wsrep_max_ws_rows= wsrep_max_ws_size= [mysqldump] quick max_allowed_packet = 128M [mysql] no-auto-rehash max_allowed_packet = 128M prompt = '(product)\u@\h:\p [\d]> ' default_character_set = utf8 [myisamchk] key_buffer_size = 64M sort_buffer_size = 512k read_buffer = 2M write_buffer = 2M [mysqlhotcopy] interactive-timeout [mysqld_safe] #malloc-lib= /usr/local/mysql/lib/mysql/libjemalloc.so
cd /usr/local/pxc/ ./scripts/mysql_install_db --defaults-file=/etc/my.cnf
看到两次ok
#修改启动脚本
vi /etc/init.d/mysql
basedir=/usr/local/pxc datadir=/data/pxc/mysql3306/data
4.启动，进行授权操作，对于第一个节点必须以特殊方式启动，如下：
/etc/init.d/mysql --help
Usage: mysql {start|stop|restart|restart-bootstrap|reload|force-reload|status|bootstrap-pxc} [ MySQL (Percona XtraDB Cluster) options ]
第一个节点启动
/etc/init.d/mysql bootstrap-pxc
5.安全加固
连接mysql：mysql -uroot -p -S /data/mysql/mysql3306/tmp/mysql.sock
delete from mysql.user where user!='root' or host!='localhost';
truncate table mysql.db;
drop database test;
进行授权，这里用户名密码是my.cnf里定义的用户名sst 密码dongzheng
GRANT ALL PRIVILEGES ON *.* TO 'sst'@'%' identified by 'dongzheng';
use mysql;
UPDATE user SET password=PASSWORD('123456') WHERE user='root';
flush privileges;
PXC状态确认
show global status like 'wsrep%';
6.netstat进行查看，可以发现启动两个端口
netstat -lntp | grep mysql
tcp 0 0 0.0.0.0:4567 0.0.0.0:* LISTEN 2964/mysqld
tcp 0 0 0.0.0.0:3306 0.0.0.0:* LISTEN 2964/mysqld
另外两个节点的启动和配置
node2 my.cnf的PXC 配置节：
############# PXC #####################
innodb_autoinc_lock_mode=2 #自增锁的优化
wsrep_cluster_name=pxc-dongzheng #集群名字
wsrep_provider=/usr/local/mysql/lib/libgalera_smm.so #库文件
wsrep_cluster_address=gcomm://192.168.2.130,192.168.3.128,192.168.0.129 #节点中所有ip
wsrep_node_address=192.168.0.129 #节点的ip
wsrep_slave_threads=2 #开启的复制线程数，cpu核数*2
wsrep_sst_auth=sst:dongzheng #sst模式需要的用户名和密码
wsrep_sst_method=xtrabackup-v2 #采用什么方式复制数据。还支持mysqldump，rsync
wsrep_provider_options="debug=1;socket.checksum=1" #打开调试模式
node3 my.cnf的PXC 配置节：
############# PXC #####################
innodb_autoinc_lock_mode=2 #自增锁的优化
wsrep_cluster_name=pxc-dongzheng #集群名字
wsrep_provider=/usr/local/mysql/lib/libgalera_smm.so #库文件
wsrep_cluster_address=gcomm://192.168.2.130,192.168.3.128,192.168.0.129 #节点中所有ip
wsrep_node_address=192.168.0.128 #节点的ip
wsrep_slave_threads=2 #开启的复制线程数，cpu核数*2
wsrep_sst_auth=sst:dongzheng #sst模式需要的用户名和密码
wsrep_sst_method=xtrabackup-v2 #采用什么方式复制数据。还支持mysqldump，rsync
wsrep_provider_options="debug=1;socket.checksum=1" #打开调试模式
启动
/etc/init.d/mysql start
启动的时候会把主节点的所有数据传送过从节点包括所有数据库和日志，简单来讲从节点在启动的时候会做SST同步
从节点不需要做安全加固，因为从节点跟主节点的数据库和数据都一样
PXC状态确认
show global status like 'wsrep%';
netstat进行查看，可以发现启动两个端口
netstat -lntp | grep mysql
tcp 0 0 0.0.0.0:4567 0.0.0.0:* LISTEN 2964/mysqld
tcp 0 0 0.0.0.0:3306 0.0.0.0:* LISTEN 2964/mysqld
###############################################################
集群关闭
所有节点都用
/etc/init.d/mysql stop
节点重启
第一个节点用：/etc/init.d/mysql restart-bootstrap
其他节点用：/etc/init.d/mysql restart
集群关闭之后，再启动，谁做第一个节点谁就执行/etc/init.d/mysql bootstrap-pxc
如果不是所有节点都关闭，那么当启动集群节点的时候不需要执行/etc/init.d/mysql bootstrap-pxc
/etc/init.d/mysql --help
Usage: mysql {start|stop|restart|restart-bootstrap|reload|force-reload|status|bootstrap-pxc} [ MySQL (Percona XtraDB Cluster) options ]
到这里搭建就基本结束了。3个节点同时支持read/write操作。
新加入节点
只需要安装好Percona-XtraDB-Cluster，并且在 my.cnf 添加PXC 配置节即可
报错处理
（1）加载 so 文件加不到
如：
Installing MySQL system tables... /usr/local/mysql/bin/mysqld: error : cannot open shared object file: No such file or directory
可能的问题：
openssl 没装：
yum install –y openssl openssl-devel
如果提示已经安装，可以使用
updatedb locate libssl.so
然后用
ldd /usr/local/mysql/bin/mysqld
查看还有那些依赖需安装
确认一下：
看样子都是存在的，但版本不对。做一个软链就可以修复了。
#cd /usr/lib64/
最后在用ldd /usr/local/mysql/bin/mysqld 去确认是否有没加载的so, 如果存在同样的方法修复。
（2）安装中 socat 找不到
/usr/local/pxc_56/bin/mysqld_safe --defaults-file=/home/mysql/my3307.cnf --ledir=/usr/local/pxc_56/bin/ &
有可能启动会提示：
WSREP_SST: [ERROR] socat not found in path: /usr/sbin:/sbin:/usr/local/pxc_56//bin:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin:/opt/lamp/mysql/bin:/root/bi
n ( ::04.404)
-- :: [ERROR] WSREP: Failed to read 'ready <addr>' from: wsrep_sst_xtrabackup-v2 --role 'joiner' --address '192.168.1.238' --datadir '/home/mysql/I3307/
'' ''
Read: '(null)'
-- :: [ERROR] WSREP: Process completed with error: wsrep_sst_xtrabackup-v2 --role 'joiner' --address '192.168.1.238' --datadir '/home/mysql/I3307/' --de
faults- (No such file or directory)
-- :: [ERROR] WSREP: Failed to prepare for 'xtrabackup-v2' SST. Unrecoverable.
-- :: [ERROR] Aborting
socat 包没装上，可以把epel 的源加入就可以安装。
yum install -y socat
因为在my.cnf里写错ip，导致出现下面错误，改正my.cnf就可以了
wsrep_cluster_address=gcomm://192.168.2.130,192.168.3.128,192.168.0.129
wsrep_node_address=192.168.1.130
-- :: [ERROR] WSREP: failed to open gcomm backend connection: : failed to reach primary view: (Connection timed out)
at gcomm/src/pc.
-- :: [ERROR] WSREP: gcs/src/gcs_core.: Failed to open backend connection: - (Connection timed out)
-- :: [ERROR] WSREP: gcs/src/gcs.: Failed to open channel (Connection timed out)
-- :: [ERROR] WSREP: gcs connect failed: Connection timed out
-- :: [ERROR] WSREP: wsrep::connect(gcomm://192.168.2.130,192.168.3.128,192.168.0.129) failed: 7
-- :: [ERROR] Aborting
关于PXC版本
如果是安装5.6.24版本或以上的需要下载openssl
centos6和centos7需要openssl版本1.0
https://www.percona.com/doc/percona-server/5.6/installation.html#installing-percona-server-from-a-binary-tarball
In Percona Server 5.6.24-72.2 and newer, the single binary tarball was replaced with multiple tarballs depending on the OpenSSL library available in the distribution:
ssl100 - for all Debian/Ubuntu versions except Debian Squeeze (libssl.so.1.0.0 => /usr/lib/x86_64-linux-gnu/libssl.so.1.0.0 (0x00007f2e389a5000));
ssl098 - only for Debian Squeeze (libssl.so.0.9.8 => /usr/lib/libssl.so.0.9.8 (0x00007f9b30db6000));
ssl101 - for CentOS 6 and CentOS 7 (libssl.so.10 => /usr/lib64/libssl.so.10 (0x00007facbe8c4000));
ssl098e - to be used only for CentOS 5 (libssl.so.6 => /lib64/libssl.so.6 (0x00002aed5b64d000)).
下载地址
https://www.percona.com/downloads/Percona-XtraDB-Cluster-56/LATEST/binary/tarball/
运维注意
（1）不能同时关闭所有节点，原则要保持Group里最少一个成员活着，否则全部传SST不能同时关闭所有节点，原则要保持Group里最少一个成员活着，否则全部传SST，gcache丢失
（2）node1 是整个集群的老大
其它节点加进来发现数据不一致，以老大为准
会有丢数据风险
（3）其中一个节点死掉了，还有2个节点
发现整个集群还能活，要设置忽略脑裂
忽略脑裂的命令
SET GLOBAL wsrep_provider_options="pc.ignore_sb=true;pc.ignore_quorum=true";
（4）最好单节点写入，不然更新同一行记录会出问题
（5）尽量用pt-online-schema-change 做表结构变更
（6）mysql库全是MyISAM引擎不能复制
MyISAM引擎不能被复制，PXC只支持Innodb
DCL语句可以复制： create user， drop user, grant ,revoke
（7）整个集群节点数最好为3，最多是8个
（8）pxc结构里面每个表必须有主键
（9）writeSet的大小
下面两个参数控制写入集大小
wsrep_max_ws_rows 没有默认值，默认没有限制
wsrep_max_ws_size 默认单位字节，建议不要超过16KB，16000
my.cnf
wsrep_max_ws_rows=131072
wsrep_max_ws_size=16000
允许最大的事务大小由wsrep_max_ws_rows和wsrep_max_ws_size定义，超过这个大小事务会被抛弃
LOAD DATA INFILE ...每1万行提交一次.对于LOAD DATA中的大事务会被分割成多个小事务来执行.
参考文章：
http://www.cnblogs.com/zejin2008/p/5475285.html
https://www.percona.com/doc/percona-xtradb-cluster/5.6/wsrep-provider-index.html
https://www.percona.com/doc/percona-xtradb-cluster/5.6/wsrep-system-index.html
http://galeracluster.com/documentation-webpages/monitoringthecluster.html
如有不对的地方，欢迎大家拍砖o(∩_∩)o
本文版权归作者所有，未经作者同意不得转载。