orchestrator 是一个MySQL高可用复制拓补的管理和可视化工具，允许：
1）发现：
orchestrator主动发现拓补结构并读取基本的MySQL信息，如复制状态和配置。
2）重构：
可以将一个不可用的服务器从拓补结构中剔除，并把数据副本移动到另一个从库下面
3）恢复：
在不可用时，可以选择一个适合的从库提升为主库
4）可视化：
web可视化，并允许直接在界面上进行复制拓补变更等操作
一、环境：
系统：
centos7.2
三台主机：
master：192.168.89.100
slave1：192.168.89.102
slave2：192.168.89.103
后端端口：
orchestrator实例端口：3306
后端mysql服务器实例端口：3307
二、安装：
orchestrator要求与mysql实例为1:1的关系。开启raft时需要关闭IPV6。关闭selinux。
1）安装yum扩展源：yum -y install epel-release
2）更新系统上的软件：yum -y update
3）安装依赖：yum install -y jq oniguruma oniguruma-devel
4）下载服务端和客户端的软件包：
wget https://github.com/github/orchestrator/releases/download/v3.0.14/orchestrator-3.0.14-1.x86_64.rpm
wget https://github.com/github/orchestrator/releases/download/v3.0.14/orchestrator-client-3.0.14-1.x86_64.rpm
5）安装：yum localinstall -y orchestrator*.rpm
6）安装后，自动在/usr/local下创建orchestrator目录。
orchestrator：执行程序。
*.conf.json：配置文件模板。
resources：web，伪GTID相关文件
7）配置文件直接使用自带的模板文件：cp -a orchestrator-sample.conf.json orchestrator.conf.json
8）配置文件参数说明：
{
"Debug": true, #debug模式，输出详细信息
"EnableSyslog": false, #是否输出到系统日志里
"ListenAddress": ":3000", #orchestrator的监听端口，web端口
"MySQLTopologyUser": "failover", #后端被管理的mysql实例中的账号，所有实例都要有，本次为3307端口
"MySQLTopologyPassword": "123456", #密码
"MySQLTopologyCredentialsConfigFile": "", #验证的配置文件，账号密码可以直接写入文件，读取
"MySQLTopologySSLPrivateKeyFile": "", #ssl验证文件
"MySQLTopologySSLCertFile": "",
"MySQLTopologySSLCAFile": "",
"MySQLTopologySSLSkipVerify": true, #跳过验证
"MySQLTopologyUseMutualTLS": false, #使用TLS验证
"MySQLOrchestratorHost": "127.0.0.1", #orchestrator的IP，也可以是本机IP
"MySQLOrchestratorPort": 3306, #orchestrator所在的端口，本次为3306端口
"MySQLOrchestratorDatabase": "orchestrator", #orchestrator元数据的数据库名称
"MySQLOrchestratorUser": "root", #管理orchestrator数据库的账户
"MySQLOrchestratorPassword": "123456", #密码
"MySQLOrchestratorCredentialsConfigFile": "",
"MySQLOrchestratorSSLPrivateKeyFile": "",
"MySQLOrchestratorSSLCertFile": "",
"MySQLOrchestratorSSLCAFile": "",
"MySQLOrchestratorSSLSkipVerify": true,
"MySQLOrchestratorUseMutualTLS": false,
"MySQLConnectTimeoutSeconds": 1, #orchestrator连接mysql超时秒数
"DefaultInstancePort": 3307, #mysql实例的端口，本次为3307，对外提供服务的实例
"DiscoverByShowSlaveHosts": true, #是否启用审查和自动发现
"InstancePollSeconds": 5, #orchestrator探测mysql间隔秒数
"SkipMaxScaleCheck": true, #没有MaxScale binlogserver设置为true
"UnseenInstanceForgetHours": 240,
"SnapshotTopologiesIntervalHours": 0,
"InstanceBulkOperationsWaitTimeoutSeconds": 10,
"HostnameResolveMethod": "none", #解析主机名，默认default 不解析为none
"MySQLHostnameResolveMethod": "@@hostname",
"SkipBinlogServerUnresolveCheck": true, #跳过二进制服务器检测
"ExpiryHostnameResolvesMinutes": 60, #域名检测过期分钟数
"RejectHostnameResolvePattern": "", #禁止的域名正则表达式
"ReasonableReplicationLagSeconds": 10, #复制延迟高于10秒表示异常
"ProblemIgnoreHostnameFilters": [], #主机正则匹配筛选最小化
"VerifyReplicationFilters": false, #重构钱检查复制筛选器
"ReasonableMaintenanceReplicationLagSeconds": 20, #上移和下移的阈值
"CandidateInstanceExpireMinutes": 60, #实例过期分钟数
"AuditLogFile": "", #审计日志
"AuditToSyslog": false, #审计日志输出到系统日志
"RemoveTextFromHostnameDisplay": ":3306", #去除集群的文本
"ReadOnly": true, #全局只读
"AuthenticationMethod": "", #身份验证模式
"HTTPAuthUser": "", #http验证用户名
"HTTPAuthPassword": "", #http验证密码
"AuthUserHeader": "",
"PowerAuthUsers": [
"*"
],
"ClusterNameToAlias": {
"127.0.0.1": "test suite"
},
"SlaveLagQuery": "", #使用SHOW SLAVE STATUS进行延迟判断
"DetectClusterAliasQuery": "SELECT SUBSTRING_INDEX(@@hostname, '.', 1)", #查询集群别名
"DetectClusterDomainQuery": "", #查询集群Domain
"DetectInstanceAliasQuery": "",
"DetectPromotionRuleQuery": "",
"DataCenterPattern": "[.]([^.]+)[.][^.]+[.]mydomain[.]com", #从正则表达式中筛选数据中心名称
"PhysicalEnvironmentPattern": "[.]([^.]+[.][^.]+)[.]mydomain[.]com", #返回实例的物理环境
"PromotionIgnoreHostnameFilters": [],
"DetectSemiSyncEnforcedQuery": "", #查询以确定是否强制完全半同步写入
"ServeAgentsHttp": false, #产生一个agent的http接口
"AgentsServerPort": ":3001",
"AgentsUseSSL": false,
"AgentsUseMutualTLS": false,
"AgentSSLSkipVerify": false,
"AgentSSLPrivateKeyFile": "",
"AgentSSLCertFile": "",
"AgentSSLCAFile": "",
"AgentSSLValidOUs": [],
"UseSSL": false,
"UseMutualTLS": false,
"SSLSkipVerify": false,
"SSLPrivateKeyFile": "",
"SSLCertFile": "",
"SSLCAFile": "",
"SSLValidOUs": [],
"URLPrefix": "",
"StatusEndpoint": "/api/status",
"StatusSimpleHealth": true,
"StatusOUVerify": false,
"AgentPollMinutes": 60,
"UnseenAgentForgetHours": 6,
"StaleSeedFailMinutes": 60,
"SeedAcceptableBytesDiff": 8192,
"PseudoGTIDPattern": "", #为空禁用伪GTID
"PseudoGTIDPatternIsFixedSubstring": false,
"PseudoGTIDMonotonicHint": "asc:",
"DetectPseudoGTIDQuery": "",
"BinlogEventsChunkSize": 10000, #查看二进制日志的块大小，更小锁定更少
"SkipBinlogEventsContaining": [],
"ReduceReplicationAnalysisCount": true,
"FailureDetectionPeriodBlockMinutes": 60, #该时间内发现故障，不被多次发现
"RecoveryPeriodBlockSeconds": 3600, #该时间内发现故障，不会多次转移
"RecoveryIgnoreHostnameFilters": [], #恢复会忽略的主机
"RecoverMasterClusterFilters": [ #对匹配的集群主恢复
"*"
],
"RecoverIntermediateMasterClusterFilters": [ #对匹配的集群恢复
"*"
],
"OnFailureDetectionProcesses": [ #故障转移之前的HOOK
"echo 'Detected {failureType} on {failureCluster}. Affected replicas: {countSlaves} autoMasterRecovery: {autoMasterRecovery} losthost: {lostSlaves} slavehost: {slaveHosts} orchestratorHost: {orchestratorHost}' >> /tmp/recovery.log"
],
"PreGracefulTakeoverProcesses": [ #主变为只读之前执行的HOOK
"echo 'Planned takeover about to take place on {failureCluster}. Master will switch to read_only autoMasterRecovery: {autoMasterRecovery} losthost: {lostSlaves} slavehost: {slaveHosts} orchestratorHost: {orchestratorHost}' >> /tmp/recovery.log"
],
"PreFailoverProcesses": [ #执行恢复操作前执行
"echo 'Will recover from {failureType} on {failureCluster}' >> /tmp/recovery.log"
],
"PostFailoverProcesses": [ #恢复成功后执行
"echo '(for all types) Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Successor: {successorHost}:{successorPort}' >> /tmp/recovery.log"
],
"PostUnsuccessfulFailoverProcesses": [], #任何失败的恢复执行
"PostMasterFailoverProcesses": [ #恢复成功时执行
"echo 'Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Promoted: {successorHost}:{successorPort}' >> /tmp/recovery.log"
],
"PostIntermediateMasterFailoverProcesses": [ #成功的中间主恢复时执行
"echo 'Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Successor: {successorHost}:{successorPort}' >> /tmp/recovery.log"
],
"PostGracefulTakeoverProcesses": [ #旧主位于新主之后执行
"echo 'Planned takeover complete' >> /tmp/recovery.log"
],
"CoMasterRecoveryMustPromoteOtherCoMaster": true, #共同恢复，否则失败
"DetachLostSlavesAfterMasterFailover": true, #恢复可能丢失一些副本
"ApplyMySQLPromotionAfterMasterFailover": true, #在将要提升的主机上设置read_only=0
"PreventCrossDataCenterMasterFailover": false, #允许跨DC转移
"MasterFailoverDetachSlaveMasterHost": false, #确保新主不复制旧主的数据
"MasterFailoverLostInstancesDowntimeMinutes": 0, #故障转移之后，其他从停机的时间
"PostponeSlaveRecoveryOnLagMinutes": 0,
"OSCIgnoreHostnameFilters": [],
"GraphiteAddr": "",
"GraphitePath": "",
"GraphiteConvertHostnameDotsToUnderscores": true,
"RaftEnabled": true, #raft模式
"BackendDB": "mysql", #后台数据库类型
"RaftBind": "192.168.89.103", #绑定之地，本机IP
"RaftDataDir": "/var/lib/orchestrator", #数据目录，如果不存在，则自动创建
"DefaultRaftPort": 10008, #raft通信端口，所有机器必须保持一致
"RaftNodes": [ #raft节点，必须包含所有节点
"192.168.89.100",
"192.168.89.102",
"192.168.89.103"
],
"ConsulAddress": "",
"ConsulAclToken": ""
}
三、启动：
1）在所有被管理的mysql实例上（本机为3307端口）建立管理账号：
GRANT ALL ON *.* TO 'failover'@'127.0.0.1' IDENTIFIED BY '123456';
GRANT ALL ON *.* TO 'failover'@'192.168.89.%' IDENTIFIED BY '123456';
在所有orchestrator实例的mysql上建立管理账号：
GRANT ALL ON *.* TO 'root'@'127.0.0.1' IDENTIFIED BY '123456';
2）创建复制拓补环境，这里不细说，把三台主机创建成复制结构即可。
3）编辑hosts文件，写入三台的hostname。
4）启动orchestrator：
nohup orchestrator http & orchestrator会自动到/usr/local/orchestrator下查找到配置文件
5）打开web界面：IP:3000。orchestrator会自动发现复制拓补结构并显示到界面上，如果没有发现，则可以手动发现。
orchestrator-client -c discover -i 实例：端口
orchestrator可以很好地与proxysql中间件结合，做读写分离与高可用。事实上，percona官方也推荐这套组合，mysqlfailover+proxysql也是很好的一套，不过MySQLfailover需要自己定制脚本。
目前，我们生产环境是orchestrator+proxysql。