目錄
Orchestrator總體結構...
測試環境信息...
Orchestrator詳細配置...
SSH免密配置...
/etc/hosts配置...
visudo配置...
/etc/orchestrator.conf.json
orch_hook.sh.
orch_vip.sh.
MySQL Server配置...
測試導換...
啟動orchestrator.
登錄Web並發現實例...
切換前...
切換后...
Orchestrator是如何監測Master異常的...
總結...
參考資料
Orchestrator是最近非常流行的MySQL復制管理工具。相關的資料也相當的多，但是總是缺點什么。
那就是步驟不夠詳細，按照博客的操作，很多都無法成功。說實話，看了N多博客后，我倒騰了一周時間才成功。
特別是Orchestrator后台也使用MySQL存儲，這樣容易把人繞進去了(分不清參數哪個是Orchestrator用，哪個是連接mysql實例用了）。
所以在這個博客，我將使用Sqlite存儲。並把每一步操作都詳細列舉出來。
Orchestrator總體結構
一圖勝千言。上圖就是Orchestrator的架構圖。
圖中：Sqlite/MySQL，是Orchestrator的后端存儲數據庫（存儲監控的mysql復制實例的相關狀態信息），可以選擇MySQL，也可以選擇SQLite。
/etc/orchestrator.conf.json是配置文件，Orchestrator啟動時讀取。
下面部分是監控的MySQL的實例，本例中，我是搭建的一主，兩從。當然，實際上Orchestrator可以監控成百上千個MySQL復制集群。
測試環境信息
orch 192.168.56.130
host01 192.168.56.103
host02 192.168.56.104
host03: 192.168.56.105
Orchestrator詳細配置
SSH免密配置
參考如下鏈接，在這里，我創建的是orch帳號。創建免費登錄的作用是shell腳本通過ssh登錄Mysql服務器進行VIP切換。
https://www.cnblogs.com/chasetimeyang/p/15064507.html
/etc/hosts配置
同樣，用於Shell腳本。
[orch@orch orchestrator]$ cat /etc/hosts 127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 ::1 localhost localhost.localdomain localhost6 localhost6.localdomain6 192.168.56.103 host01 192.168.56.104 host02 192.168.56.105 host03 192.168.56.130 orch
visudo配置
因為我的VIP shell導換腳本內，用到了sudo執行高權限命令。你可以通過如下方法把用戶加入sudo權限。
$visudo
orch ALL=(ALL) NOPASSWD: ALL
/etc/orchestrator.conf.json
如下兩個參數是連接要監控的MySQL實例的連接用戶名和密碼
"MySQLTopologyUser": "orchestrator",
"MySQLTopologyPassword": "orc_topology_password",
如下說明是Sqlite作為后端存儲數據庫
"BackendDB": "sqlite",
"SQLite3DataFile": "/usr/local/orchestrator/orchestrator.sqlite3",
如下參數是實例導換后，多長時間內不允許再次導換。默認是60分鍾。
"FailureDetectionPeriodBlockMinutes": 5,
如下參數是指，恢復匹配任何實例。（你也可以配置匹配規則，如此滿足規則是會自動導換。不滿足的，出現問題時，只能手動導換）
"RecoverMasterClusterFilters": [
"*"
],
"RecoverIntermediateMasterClusterFilters": [
"*"
],
是否自動導換。
"ApplyMySQLPromotionAfterMasterFailover": true,
這個是導換后執行的腳本
"PostFailoverProcesses": [
"echo '(for all types) Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Successor: {successorHost}:{successorPort}' >> /tmp/recovery.log",
"/home/orch/orch_hook.sh {failureType} {failureClusterAlias} {failedHost} {successorHost} >> /tmp/orch.log"
],
[orch@orch ~]$ cat /etc/orchestrator.conf.json { "Debug": true, "EnableSyslog": false, "ListenAddress": ":3000", "MySQLTopologyUser": "orchestrator", "MySQLTopologyPassword": "orc_topology_password", "MySQLTopologyCredentialsConfigFile": "", "MySQLTopologySSLPrivateKeyFile": "", "MySQLTopologySSLCertFile": "", "MySQLTopologySSLCAFile": "", "MySQLTopologySSLSkipVerify": true, "MySQLTopologyUseMutualTLS": false, "BackendDB": "sqlite", "SQLite3DataFile": "/usr/local/orchestrator/orchestrator.sqlite3", "MySQLConnectTimeoutSeconds": 1, "DefaultInstancePort": 3306, "DiscoverByShowSlaveHosts": true, "InstancePollSeconds": 5, "DiscoveryIgnoreReplicaHostnameFilters": [ "a_host_i_want_to_ignore[.]example[.]com", ".*[.]ignore_all_hosts_from_this_domain[.]example[.]com", "a_host_with_extra_port_i_want_to_ignore[.]example[.]com:3307" ], "UnseenInstanceForgetHours": 240, "SnapshotTopologiesIntervalHours": 0, "InstanceBulkOperationsWaitTimeoutSeconds": 10, "HostnameResolveMethod": "default", "MySQLHostnameResolveMethod": "@@hostname", "SkipBinlogServerUnresolveCheck": true, "ExpiryHostnameResolvesMinutes": 60, "RejectHostnameResolvePattern": "", "ReasonableReplicationLagSeconds": 10, "ProblemIgnoreHostnameFilters": [], "VerifyReplicationFilters": false, "ReasonableMaintenanceReplicationLagSeconds": 20, "CandidateInstanceExpireMinutes": 1, "AuditLogFile": "", "AuditToSyslog": false, "RemoveTextFromHostnameDisplay": ".mydomain.com:3306", "ReadOnly": false, "AuthenticationMethod": "", "HTTPAuthUser": "", "HTTPAuthPassword": "", "AuthUserHeader": "", "PowerAuthUsers": [ "*" ], "ClusterNameToAlias": { "127.0.0.1": "test suite" }, "ReplicationLagQuery": "", "DetectClusterAliasQuery": "SELECT SUBSTRING_INDEX(@@hostname, '.', 1)", "DetectClusterDomainQuery": "", "DetectInstanceAliasQuery": "", "DetectPromotionRuleQuery": "", "DataCenterPattern": "[.]([^.]+)[.][^.]+[.]mydomain[.]com", "PhysicalEnvironmentPattern": "[.]([^.]+[.][^.]+)[.]mydomain[.]com", "PromotionIgnoreHostnameFilters": [], "DetectSemiSyncEnforcedQuery": "", "ServeAgentsHttp": false, "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": "", "PseudoGTIDPatternIsFixedSubstring": false, "PseudoGTIDMonotonicHint": "asc:", "DetectPseudoGTIDQuery": "", "BinlogEventsChunkSize": 10000, "SkipBinlogEventsContaining": [], "ReduceReplicationAnalysisCount": true, "FailureDetectionPeriodBlockMinutes": 5, "FailMasterPromotionOnLagMinutes": 0, "RecoveryPeriodBlockSeconds": 3600, "RecoveryIgnoreHostnameFilters": [], "RecoverMasterClusterFilters": [ "*" ], "RecoverIntermediateMasterClusterFilters": [ "*" ], "OnFailureDetectionProcesses": [ "echo 'Detected {failureType} on {failureCluster}. Affected replicas: {countSlaves}' >> /tmp/recovery.log" ], "PreGracefulTakeoverProcesses": [ "echo 'Planned takeover about to take place on {failureCluster}. Master will switch to read_only' >> /tmp/recovery.log" ], "PreFailoverProcesses": [ "echo 'Will recover from {failureType} on {failureCluster}' >> /tmp/recovery.log" ], "PostFailoverProcesses": [ "echo '(for all types) Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Successor: {successorHost}:{successorPort}' >> /tmp/recovery.log", "/home/orch/orch_hook.sh {failureType} {failureClusterAlias} {failedHost} {successorHost} >> /tmp/orch.log" ], "PostUnsuccessfulFailoverProcesses": [], "PostMasterFailoverProcesses": [ "echo 'Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Promoted: {successorHost}:{successorPort}' >> /tmp/recovery.log" ], "PostIntermediateMasterFailoverProcesses": [ "echo 'Recovered from {failureType} on {failureCluster}. Failed: {failedHost}:{failedPort}; Successor: {successorHost}:{successorPort}' >> /tmp/recovery.log" ], "PostGracefulTakeoverProcesses": [ "echo 'Planned takeover complete' >> /tmp/recovery.log" ], "CoMasterRecoveryMustPromoteOtherCoMaster": true, "DetachLostSlavesAfterMasterFailover": true, "ApplyMySQLPromotionAfterMasterFailover": true, "PreventCrossDataCenterMasterFailover": false, "PreventCrossRegionMasterFailover": false, "MasterFailoverDetachReplicaMasterHost": false, "MasterFailoverLostInstancesDowntimeMinutes": 0, "PostponeReplicaRecoveryOnLagMinutes": 0, "OSCIgnoreHostnameFilters": [], "GraphiteAddr": "", "GraphitePath": "", "GraphiteConvertHostnameDotsToUnderscores": true, "ConsulAddress": "", "ConsulAclToken": "", "ConsulKVStoreProvider": "consul" }
orch_hook.sh
#!/bin/bash
isitdead=$1
#cluster=$2
oldmaster=$3
newmaster=$4
ssh=$(which ssh)
logfile="/home/orch/orch_hook.log"
interface='enp0s3'
user=orch
#VIP=$($ssh -tt ${user}@${oldmaster} "sudo ip address show dev enp0s3|grep -w 'inet'|tail -n 1|awk '{print \$2}'|awk -F/ '{print \$1}'")
VIP='192.168.56.200'
VIP_TEMP=$($ssh -tt ${user}@${oldmaster} "sudo ip address|sed -nr 's#^.*inet (.*)/32.*#\1#gp'")
#remove '\r' at the end $'192.168.56.200\r'
VIP_TEMP=$(echo $VIP_TEMP|awk -F"\\r" '{print $1}')
if [ ${#VIP_TEMP} -gt 0 ]; then
VIP=$VIP_TEMP
fi
echo ${VIP}
echo ${interface}
if [[ $isitdead == "DeadMaster" ]]; then
if [ !-z ${!VIP} ] ; then
echo $(date)
echo "Revocering from: $isitdead"
echo "New master is: $newmaster"
echo "/home/orch/orch_vip.sh -d 1 -n $newmaster -i ${interface} -I $VIP -u ${user} -o ${oldmaster}"
/home/orch/orch_vip.sh -d 1 -n $newmaster -i ${interface} -I $VIP -u ${user} -o ${oldmaster}
else
echo "Cluster does not exist!" | tee $logfile
fi
fi
orch_vip.sh
#!/bin/bash
function usage {
cat << EOF
usage: $0 [-h] [-d master is dead] [-o old master ] [-s ssh options] [-n new master] [-i interface] [-I] [-u SSH user]
OPTIONS:
-h Show this message
-o string Old master hostname or IP address
-d int If master is dead should be 1 otherweise it is 0
-s string SSH options
-n string New master hostname or IP address
-i string Interface exmple eth0:1
-I string Virtual IP
-u string SSH user
EOF
}
while getopts ho:d:s:n:i:I:u: flag; do
case $flag in
o)
orig_master="$OPTARG";
;;
d)
isitdead="${OPTARG}";
;;
s)
ssh_options="${OPTARG}";
;;
n)
new_master="$OPTARG";
;;
i)
interface="$OPTARG";
;;
I)
vip="$OPTARG";
;;
u)
ssh_user="$OPTARG";
;;
h)
usage;
exit 0;
;;
*)
usage;
exit 1;
;;
esac
done
if [ $OPTIND -eq 1 ]; then
echo "No options were passed";
usage;
fi
shift $(( OPTIND - 1 ));
# discover commands from our path
ssh=$(which ssh)
arping=$(which arping)
ip2util=$(which ip)
#ip2util='ip'
# command for adding our vip
cmd_vip_add="sudo -n $ip2util address add $vip dev $interface"
# command for deleting our vip
cmd_vip_del="sudo -n $ip2util address del $vip/32 dev $interface"
# command for discovering if our vip is enabled
cmd_vip_chk="sudo -n $ip2util address show dev $interface to ${vip%/*}/32"
# command for sending gratuitous arp to announce ip move
cmd_arp_fix="sudo -n $arping -c 1 -I ${interface} ${vip%/*}"
# command for sending gratuitous arp to announce ip move on current server
#cmd_local_arp_fix="sudo -n $arping -c 1 -I ${interface} ${vip%/*}"
cmd_local_arp_fix="$arping -c 1 -I ${interface} ${vip%/*}"
vip_stop() {
rc=0
echo $?
echo "$ssh ${ssh_options} -tt ${ssh_user}@${orig_master} \
\"[ -n \"\$(${cmd_vip_chk})\" ] && ${cmd_vip_del} && \
sudo -n ${ip2util} route flush cache || [ -z \"\$(${cmd_vip_chk})\" ]\""
# ensure the vip is removed
$ssh ${ssh_options} -tt ${ssh_user}@${orig_master} \
"[ -n \"\$(${cmd_vip_chk})\" ] && ${cmd_vip_del} && \
sudo -n ${ip2util} route flush cache || [ -z \"\$(${cmd_vip_chk})\" ]"
rc=$?
return $rc
}
vip_start() {
rc=0
# ensure the vip is added
# this command should exit with failure if we are unable to add the vip
# if the vip already exists always exit 0 (whether or not we added it)
echo "$ssh ${ssh_options} -tt ${ssh_user}@${new_master} \
\"[ -z \"\$(${cmd_vip_chk})\" ] && ${cmd_vip_add} && ${cmd_arp_fix} || [ -n \"\$(${cmd_vip_chk})\" ]\""
$ssh ${ssh_options} -tt ${ssh_user}@${new_master} \
"[ -z \"\$(${cmd_vip_chk})\" ] && ${cmd_vip_add} && ${cmd_arp_fix} || [ -n \"\$(${cmd_vip_chk})\" ]"
rc=$?
echo "vip started"
#$cmd_local_arp_fix
return $rc
}
vip_status() {
$arping -c 1 -I ${interface} ${vip%/*}
echo "$arping -c 1 -I ${interface} ${vip%/*}"
if ping -c 1 -W 1 "$vip"; then
return 0
else
return 1
fi
}
if [[ $isitdead == 0 ]]; then
echo "Online failover"
if vip_stop; then
if vip_start; then
echo "$vip is moved to $new_master."
else
echo "Can't add $vip on $new_master!"
exit 1
fi
else
echo $rc
echo "Can't remove the $vip from orig_master!"
exit 1
fi
elif [[ $isitdead == 1 ]]; then
echo "Master is dead, failover"
# make sure the vip is not available
if vip_status; then
if vip_stop; then
echo "$vip is removed from orig_master."
else
echo $rc
echo "Couldn't remove $vip from orig_master."
exit 1
fi
fi
if vip_start; then
echo "$vip is moved to $new_master."
else
echo "Can't add $vip on $new_master!"
exit 1
fi
else
echo "Wrong argument, the master is dead or live?"
fi
MySQL Server配置
配置一下Master和不同的Slave的配置文件 (/etc/my.cnf)。所有配置參數可以一樣，除了server_id。
需要注意的是每個MySQL的server_id必須不同。
[mysqld] datadir=/var/lib/mysql socket=/var/lib/mysql/mysql.sock log-error=/var/lib/mysql/mysqld.log pid-file=/var/lib/mysql/mysqld.pid sql_mode=NO_ENGINE_SUBSTITUTION,STRICT_TRANS_TABLES port = 3306 #GTID: server_id=135 #服務器id gtid_mode=on #開啟gtid模式 enforce_gtid_consistency=on #強制gtid一致性，開啟后對於特定create table不被支持 #binlog log_bin=binlog log-slave-updates=1 binlog_format=row #強烈建議，其他格式可能造成數據不一致 #relay log skip_slave_start=1 max_connect_errors=1000 default_authentication_plugin = 'mysql_native_password' slave_net_timeout = 4
從機連接主機，並開啟復制
mysql>change master to master_host='master', master_port=3306, master_user='repl', master_password='Xiaopang*803', master_auto_position=1,MASTER_HEARTBEAT_PERIOD=2,MASTER_CONNECT_RETRY=1, MASTER_RETRY_COUNT=86400; mysql>start slave;
復制啟動無誤后。在Master上執行如下命令創建orchestrator用戶並賦予權限(這就是上面Orachestrator用來連接MySQL復制實例的用戶）
CREATE USER 'orchestrator'@'%' IDENTIFIED BY 'orc_topology_password'; GRANT SUPER, PROCESS, REPLICATION SLAVE, REPLICATION CLIENT, RELOAD ON *.* TO 'orchestrator'@'%';
測試導換
啟動orchestrator
[orch@orch orchestrator]$ pwd
/usr/local/orchestrator
[root@orch orchestrator]# ./orchestrator --debug --config=/etc/orchestrator.conf.json http
登錄Web並發現實例
我的orchestrator的IP地址是130。
http://192.168.56.130:3000/
發現新的實例是，填上當前的Master的實例IP與端口。
切換前
手動給host03增加一個VIP地址 192.168.56.200
[root@host03 ~]# ip addr add dev enp0s3 192.168.56.200/32
查看一下
[root@host03 ~]# ip addr show dev enp0s3
enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 08:00:27:9a:25:16 brd ff:ff:ff:ff:ff:ff
inet 192.168.56.105/24 brd 192.168.56.255 scope global enp0s3
valid_lft forever preferred_lft forever
inet 192.168.56.200/32 scope global enp0s3
valid_lft forever preferred_lft forever
inet6 fe80::bfdd:5d6a:d4f9:6f95/64 scope link
valid_lft forever preferred_lft forever
inet6 fe80::5a97:c3c9:a9df:466f/64 scope link tentative dadfailed
valid_lft forever preferred_lft forever
切換后
停止host03 MySQL
[root@host03 ~]#service mysqld stop
停止后，過了1分多鍾，服務發了導換。新的拓撲結構變為如下
查看VIP是否也發生了切換。如下可以看到新的VIP已經切換到了新的主機host01上。
[root@host01 ~]# ip addr show dev enp0s3 2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 08:00:27:28:0d:72 brd ff:ff:ff:ff:ff:ff inet 192.168.56.103/24 brd 192.168.56.255 scope global enp0s3 valid_lft forever preferred_lft forever inet 192.168.56.200/32 scope global enp0s3 valid_lft forever preferred_lft forever inet6 fe80::8428:fc7:68fc:1079/64 scope link valid_lft forever preferred_lft forever inet6 fe80::bfdd:5d6a:d4f9:6f95/64 scope link tentative dadfailed valid_lft forever preferred_lft forever inet6 fe80::5a97:c3c9:a9df:466f/64 scope link tentative dadfailed valid_lft forever preferred_lft forever [root@host03 ~]# ip addr show dev enp0s3 2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 08:00:27:9a:25:16 brd ff:ff:ff:ff:ff:ff inet 192.168.56.105/24 brd 192.168.56.255 scope global enp0s3 valid_lft forever preferred_lft forever inet6 fe80::bfdd:5d6a:d4f9:6f95/64 scope link valid_lft forever preferred_lft forever inet6 fe80::5a97:c3c9:a9df:466f/64 scope link tentative dadfailed valid_lft forever preferred_lft forever
Orchestrator是如何監測Master異常的
Orchestrator不僅僅只依靠Master獨立來判斷的。實際上它還會聯系各個Slave。
當Slave達成共識Master掛掉了時，這時Orchestrator才會發生導換。
如此能夠保證更高的可靠性。
Failure detection
orchestrator uses a holistic approach to detect master and intermediate master failures.
In a naive approach, a monitoring tool would probe the master, for example, and alert when it cannot contact or query the master server. Such approach is susceptible to false positives caused by network glitches. The naive approach further mitigates false positives by running n tests spaced by t-long intervals. This reduces, in some cases, the chances of false positives, and then increases the response time in the event of real failure.
orchestrator harnesses the replication topology. It observes not only the server itself, but also its replicas. For example, to diagnose a dead master scenario, orchestrator must both:
- Fail to contact said master
- Be able to contact master's replicas, and confirm they, too, cannot see the master.
Instead of triaging the error by time, orchestrator triages by multiple observers, the replication topology servers themselves. In fact, when all of a master's replicas agree that they cannot contact their master, the replication topology is broken de-facto, and a failover is justified.
orchestrator's holistic failure detection approach is known to be very reliable in production.
總結
1）orch_hook腳本是從網上找的。相對於自己的環境改了一些地方。
這個肯定不完善，會存在一些bug。這個實驗的另一目的也是驗證Orchestrator的hook腳本的工作是否有效。
2）通過自己寫腳本的方案，感覺還是容易出問題。其實，在有現成的更好的方案下，盡量不要自己寫。
比如說，我覺得采用半同步+keepbyalive實現VIP的方案可能更好。一個Master兩個Slave，一個配置為半同步(SlaveSync)，其它的從結點配置為異步。
Master和SlaveSync，通過keepbyalive實現VIP切換(Orchestrator Hook啟停keepbyalive服務實現VIP切換）
3）網上說有更佳的辦法通過ProxySQL+Orchestartor，下次試一下。
參考資料
列出部分參考資料
https://www.cnblogs.com/zhoujinyi/p/10394389.html
https://www.jianshu.com/p/91833222581a
https://github.com/openark/orchestrator/