我整个周末都在绞尽脑汁尝试使用 CentOS 7 上的 MariaDB 10.1.5（包括 Galera 25.3.9）启动并运行 Galera 集群。该服务器运行两个数据库实例，一个仅本地，一个作为集群的一部分。
我已经多次完成此精确设置，没有任何问题，但现在我却一无所知......
我的.cnf：
[mysqld_multi]
mysqld=/usr/bin/mysqld_safe --basedir=/usr
user=mysql
#
# Server local DB
#
[mysqld1]
mysqld=/usr/bin/mysqld_safe --basedir=/usr
user=mysql
port=3307
pid-file=/storage/local/mysqld.pid
socket=/storage/local/mysqld.sock
datadir=/storage/local
log-error=/var/log/mariadb-local.log
#
# Server cluster DB
#
[mysqld2]
wsrep_new_cluster
wsrep_cluster_address=gcomm://
#wsrep_cluster_address="gcomm://172.16.1.91,172.16.1.94,172.16.1.95"
wsrep_debug=1
query_cache_size=0
log-bin=/var/log/mariadb/cluster-bin.log
binlog_format=ROW
default_storage_engine=innodb
innodb_autoinc_lock_mode=2
wsrep_provider=/usr/lib64/galera/libgalera_smm.so
wsrep_cluster_name="MyApps"
wsrep_node_address=172.16.1.91
wsrep_node_name="dbnode1"
wsrep_sst_method=xtrabackup-v2
wsrep_sst_auth=sstuser:sstpass
mysqld=/usr/bin/mysqld_safe --basedir=/usr
user=mysql
port=3306
pid-file=/storage/cluster/mysqld.pid
socket=/storage/cluster/mysqld.sock
datadir=/storage/cluster
log-error=/var/log/mariadb-cluster.log
启动此节点时我在日志中收到此信息：
2015-08-02 20:52:52 mysqld_safe Starting mysqld daemon with databases from /storage/cluster
2015-08-02 20:52:52 140344148072576 [Note] /usr/sbin/mysqld (mysqld 10.1.5-MariaDB-log) starting as process 11301 ...
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Using mutexes to ref count buffer pool pages
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: The InnoDB memory heap is disabled
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Memory barrier is not used
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Compressed tables use zlib 1.2.7
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Using Linux native AIO
2015-08-02 20:52:52 140344148072576 [Note] InnoDB: Not using CPU crc32 instructions
2015-08-02 20:52:53 140344148072576 [Note] InnoDB: Initializing buffer pool, size = 128.0M
2015-08-02 20:52:53 140344148072576 [Note] InnoDB: Completed initialization of buffer pool
2015-08-02 20:52:53 140344148072576 [Note] InnoDB: Highest supported file format is Barracuda.
2015-08-02 20:52:53 140344148072576 [Note] Plugin 'FEEDBACK' is disabled.
2015-08-02 20:52:53 140233078593664 [Note] Recovering after a crash using tc.log
2015-08-02 20:52:53 140233078593664 [Note] Starting crash recovery...
2015-08-02 20:52:53 140233078593664 [Note] Crash recovery finished.
2015-08-02 20:52:53 140232389265152 [Note] InnoDB: not started
2015-08-02 20:52:53 140344148072576 [Note] Server socket created on IP: '::'.
2015-08-02 20:52:53 140344148072576 [Note] Event Scheduler: Loaded 0 events
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Setting wsrep_ready to 0
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Read WSREPXid from InnoDB: 00000000-0000-0000-0000-000000000000:-1
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Read nil XID from storage engines, skipping position init
2015-08-02 20:52:53 140344148072576 [Note] WSREP: wsrep_load(): loading provider library '/usr/lib64/galera/libgalera_smm.so'
2015-08-02 20:52:53 140344148072576 [Note] WSREP: wsrep_load(): Galera 25.3.9(r3387) by Codership Oy <[email protected]> loaded successfully.
2015-08-02 20:52:53 140344148072576 [Note] WSREP: CRC-32C: using "slicing-by-8" algorithm.
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Found saved state: 00000000-0000-0000-0000-000000000000:-1
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Passing config to GCS: base_host = 172.16.1.91; base_port = 4567; cert.log_conflicts = no; debug = no; evs.auto_evict = 0; evs.delay_margin = PT1S; evs.delayed_keep_period = PT30S; evs.inactive_check_period = PT0.5S; evs.inactive_timeout = PT15S; evs.join_retrans_period = PT1S; evs.max_install_timeouts = 3; evs.send_window = 4; evs.stats_report_period = PT1M; evs.suspect_timeout = PT5S; evs.user_send_window = 2; evs.view_forget_timeout = PT24H; gcache.dir = /storage/cluster/; gcache.keep_pages_size = 0; gcache.mem_size = 0; gcache.name = /storage/cluster//galera.cache; gcache.page_size = 128M; gcache.size = 128M; gcs.fc_debug = 0; gcs.fc_factor = 1.0; gcs.fc_limit = 16; gcs.fc_master_slave = no; gcs.max_packet_size = 64500; gcs.max_throttle = 0.25; gcs.recv_q_hard_limit = 9223372036854775807; gcs.recv_q_soft_limit = 0.25; gcs.sync_donor = no; gmcast.segment = 0; gmcast.version = 0; pc.announce_timeout = PT3S; pc.checksum = false; pc.ignore_quorum = false; pc.ignore_sb = false; pc.npvo = false; pc.reco
2015-08-02 20:52:53 140343261148928 [Note] WSREP: Service thread queue flushed.
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Assign initial position for certification: -1, protocol version: -1
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Start replication
2015-08-02 20:52:53 140344148072576 [Note] WSREP: 'wsrep-new-cluster' option used, bootstrapping the cluster
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Setting initial position to 00000000-0000-0000-0000-000000000000:-1
2015-08-02 20:52:53 140344148072576 [Note] WSREP: protonet asio version 0
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Using CRC-32C for message checksums.
2015-08-02 20:52:53 140344148072576 [Note] WSREP: backend: asio
2015-08-02 20:52:53 140344148072576 [Warning] WSREP: access file(gvwstate.dat) failed(No such file or directory)
2015-08-02 20:52:53 140344148072576 [Note] WSREP: restore pc from disk failed
2015-08-02 20:52:53 140344148072576 [Note] WSREP: GMCast version 0
2015-08-02 20:52:53 140344148072576 [Note] WSREP: (71c20f80, 'tcp://0.0.0.0:4567') listening at tcp://0.0.0.0:4567
2015-08-02 20:52:53 140344148072576 [Note] WSREP: (71c20f80, 'tcp://0.0.0.0:4567') multicast: , ttl: 1
2015-08-02 20:52:53 140344148072576 [Note] WSREP: EVS version 0
2015-08-02 20:52:53 140344148072576 [Note] WSREP: gcomm: bootstrapping new group 'MyApps'
2015-08-02 20:52:53 140344148072576 [Note] WSREP: start_prim is enabled, turn off pc_recovery
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Node 71c20f80 state prim
2015-08-02 20:52:53 140344148072576 [Note] WSREP: view(view_id(PRIM,71c20f80,1) memb {
71c20f80,0
} joined {
} left {
} partitioned {
})
2015-08-02 20:52:53 140344148072576 [Note] WSREP: save pc into disk
2015-08-02 20:52:53 140344148072576 [Note] WSREP: gcomm: connected
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Changing maximum packet size to 64500, resulting msg size: 32636
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Shifting CLOSED -> OPEN (TO: 0)
2015-08-02 20:52:53 140344148072576 [Note] WSREP: Opened channel 'MyApps'
2015-08-02 20:52:53 140343215384320 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 0, memb_num = 1
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Starting new group from scratch: 71c26ca7-3958-11e5-aa0e-92c6feb108bf
2015-08-02 20:52:53 140343215384320 [Note] WSREP: STATE_EXCHANGE: sent state UUID: 71c2791f-3958-11e5-9fcc-52d3a70422f3
2015-08-02 20:52:53 140343215384320 [Note] WSREP: STATE EXCHANGE: sent state msg: 71c2791f-3958-11e5-9fcc-52d3a70422f3
2015-08-02 20:52:53 140343215384320 [Note] WSREP: STATE EXCHANGE: got state msg: 71c2791f-3958-11e5-9fcc-52d3a70422f3 from 0 (dbnode1)
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Quorum results:
version = 3,
component = PRIMARY,
conf_id = 0,
members = 1/1 (joined/total),
act_id = 0,
last_appl. = -1,
protocols = 0/7/3 (gcs/repl/appl),
group UUID = 71c26ca7-3958-11e5-aa0e-92c6feb108bf
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Flow-control interval: [16, 16]
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Restored state OPEN -> JOINED (0)
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Member 0.0 (dbnode1) synced with group.
2015-08-02 20:52:53 140343215384320 [Note] WSREP: Shifting JOINED -> SYNCED (TO: 0)
2015-08-02 20:52:53 140344147060480 [Note] WSREP: New cluster view: global state: 71c26ca7-3958-11e5-aa0e-92c6feb108bf:0, view# 1: Primary, number of nodes: 1, my index: 0, protocol version 3
2015-08-02 20:52:53 140344148072576 [Note] /usr/sbin/mysqld: ready for connections.
这表明一切都很好，但检查 wsrep 变量表明WSREP_READY = OFF，并且任何访问引擎的尝试都会返回
ERROR 1047 (08S01): WSREP has not yet prepared node for application use
日志看起来很好，没有错误（除了崩溃恢复，这是因为关机时间太长，可能是因为有东西挂了），唯一让我担心的是
WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 0, memb_num = 1
我觉得很奇怪，我正在引导集群，不应该这样吗bootstrap = yes？
答案1
mariadb 10.1 附带的 systemd 单元文件尚不支持引导 galera。
您可以使用以下解决方法：
[root@galera01 mysql]# systemctl set-environment MYSQLD_OPTS="--wsrep-new-cluster"
[root@galera01 mysql]# systemctl start mariadb
[root@galera01 mysql]# systemctl unset-environment MYSQLD_OPTS
这是跟踪该问题的 mariadb 错误：https://mariadb.atlassian.net/browse/MDEV-7752
答案2
这可能不是你期望的解决方案，但你有没有考虑过使用Percona XtraDB集群? 它是与 Gallera 库捆绑并配置的 Percona Server（MySQL Server 的完全兼容替代品）。
如果你严格玛拉雅数据库，您始终可以使用默认的 Percona Server 配置文件作为参考。