需求:為了容災,需要部署兩台服務器,便於主備切換,當主服務器掛掉之后,主輔服務器角色永遠對換,即使之前主服務器恢復也不再重新恢復之前主輔角色,直到下次發生主備切換
分析:mongodb兩種數據復制方式:1、主備,當主服務器掛掉之后,輔服務器不能夠提供寫服務,故mongodb主備方案不合適。2、replica set,但只有兩個節點時,主節點掛掉,輔節點也不能提供寫服務,與主備效果一樣,所以至少需要三個mongodb實例,但當前只有兩台服務器,所以有一台服務器要安裝有兩個mongodb實例,一台機器上不需要兩份數據拷貝,故第三個節點使用arbiter,只進行投票,不會被選為主節點,也不會拷貝數據。另外考慮到需求中主備邏輯,結合replica set特點,arbiter只能在服務器中運行,考慮到主備角色調換,故兩台服務器每台需要安裝兩個mongodb實例,其中一個為arbiter
兩台服務器:
192.168.103.5 (原始主服務器)
192.168.103.28 (原始輔服務器)
一、兩台服務器分別安裝mongodb
安裝步驟可參考:https://www.cnblogs.com/qq931399960/p/10425803.html
二、兩台服務器分別安裝arbiter
1、新建mongodb日志和數據庫存儲路徑
mkdir -p /home/mongo-arbiter/db mkdir -p /home/mongo-arbiter/log
2、拷貝配置文件
cp /etc/mongod.conf /etc/mongod-arbiter.conf
3、修改配置文件
vim /etc/mongod-arbiter.conf
systemLog.path修改為/home/mongo-arbiter/log/mongod.log
Storage.dbPath修改為/home/mongo-arbiter/db
processManagement.pidFilePath修改為/var/run/mongodb/mongod-arbiter.pid
net.port修改為27018
4、修改權限
chown -R mongod:mongod /home/mongo-arbiter
5、設置arbiter開機啟動
cp /usr/lib/systemd/system/mongod.service /usr/lib/systemd/system/mongod-arbiter.service
修改開機啟動文件
vim /usr/lib/systemd/system/mongod-arbiter.service
Environment改為:"OPTIONS=-f /home/mongo-arbiter/mongod.conf"
PIDFile改為:/var/run/mongodb/mongod-arbiter.pid
使配置生效
systemctl enable mongod-arbiter.service
systemctl daemon-reload
初始化用戶(四個實例的用戶要一致)
cat ./mongdb/initUser.js | mongo --port 27018 --shell
6、登錄檢測
mongo -uroot -pabc123 --authenticationDatabase "admin" --port 27018
正常登陸,ok
三、配置replica set
1、停止所有mongodb實例
2、生成keyfile文件
openssl rand -base64 756 > /home/replSetKey chmod 400 /home/replSetKey chown -R mongod:mongod /home/replSetKey
2、將上述文件拷貝到另外一台服務器/home下(路徑可自定義,只要四個實例使用相同kefile文件即可)
3、添加/修改mongodb實例配置(如下配置每個實例相同,4個mongodb實例都需要修改,如下列出的只有新加和修改的,不變的未列出)
security: keyFile: /home/replSetKey ## 新加 replication: replSetName: mongoSet ## 新加 net: bindIp: 0.0.0.0 ## 修改
4、啟動主服務器默認27017端口mongodb實例,啟動輔服務器中兩個mongodb實例
systemctl start mongod
systemctl start mongod-arbiter
5、查看防火牆是否關閉,或者27017和27018端口是否放開
未關閉防火牆或者以上端口不能訪問,則初始化replica set失敗
關閉防火牆命令
systemctl stop firewalld
systemctl disable firewalld
6、在主服務器(103.5)進入mongodb命令
mongo -u "root" -p "abc123" --authenticationDatabase "admin"
7、初始化兩個memeber的replicat set
config = {_id:"mongoSet", members:[{_id:0,host:"192.168.103.5:27017",priority:2}, {_id:1,host:"192.168.103.28:27017",priority:1}]} rs.initiate(config, true)
此時可通過rs.status()狀態
8、添加arbiter仲裁節點
rs.addArb("192.168.103.28:27018")
rs.status()查看各節點狀態,如果有節點stateStr顯示not reachable/healthy,health顯示為0,則重啟該節點,再次通過rs.status()查看是否正常
到此,主備服務基本配置結束
java客戶端連接地址為:
mongodb://root:abc123@192.168.103.5:27017,192.168.103.28:27017/?authSource=mydb&replicaSet=mongoSet
四、重新配置replica set
當主備服務器切換,java客戶端連接由於配置了replicaSet,可以自動去新主節點進行業務處理,不會有問題,當主設備恢復為后,原主服務器mongodb再次成為了主節點,但其他服務已經在新主服務器上
A、mongodb和web服務不在台機器,中間有網絡傳輸,影響性能
B、此時的replica set中,主節點在輔服務器上,主服務器上只有輔節點和arbiter,當主服務器掛掉,只有一個mongodb節點,不能在提供寫服務
綜上兩點,尤其是第二點,需要重新配置replica set,登錄到mongodb主節點
mongo -uroot -pabc123 --authenticationDatabase "admin"
或者在主服務器遠程登錄恢復的mongodb主節點
mongo -u "root" -p "abc123" --authenticationDatabase "admin" --host 192.168.103.5
重新配置
reconfig = {_id:"mongoSet","protocolVersion" : 1, members:[{_id:0,host:"192.168.103.5:27017",priority:1}, {_id:1,host:"192.168.103.28:27017",priority:2}]} rs.reconfig(reconfig, true)
添加arbiter節點
rs.addArb("192.168.103.5:27018")
使用rs.status()查看狀態,如果有節點stateStr顯示not reachable/healthy,health顯示為0,則重啟該節點,再次通過rs.status()查看是否正常
PS:
1、若有節點stateStr出現not reachable/healthy,health顯示為0,則該節點可能是arbiter節點,官方不推薦將arbiter節點安裝到存在primary和slave節點的服務器上,不過當前只有兩台服務器,也只有這樣了,可以正常運行。
2、重新配置時,需要保證之前已存在的host的id是相同的
3、重新配置需要添加protocolVersion
五、使用腳本簡化操作
1、提前准備好配置文件和arbiter開機啟動service文件,需要的字段都添加上,但replica set相關字段為注釋狀態,因為打開后,所有節點都將無法進行寫操作,新建用戶后打開,並重啟mongodb
2、提前准備好kefile,兩台服務器都使用該文件
3、安裝mongodb (數據庫和日志路徑可以自定義)

#!/bin/bash basedir=$(cd "$(dirname "$0")";pwd) function initenv(){ setenforce 0 sed -i 's/SELINUX=enforcing/SELINUX=disabled/g' /etc/selinux/config } ########inst mongdb 4.0############# function inst(){ cp -f ./mongdb/mongod.conf /etc/ rpm -ivh ./mongdb/*rpm cd /home mkdir -p mongo-db mkdir -p mongo-log chown -R mongod:mongod ./mongo-db chown -R mongod:mongod ./mongo-log systemctl enable mongod systemctl start mongod } function initdb(){ cd $basedir cat ./mongdb/initUser.js | mongo --shell cat ./mongdb/initIndex.js | mongo -u "root" -p "abc123" --authenticationDatabase "admin" } function cpReplicaSetKey(){ cd $basedir cp ./mongdb/replSetKey /home/ chmod 400 /home/replSetKey chown -R mongod:mongod /home/replSetKey } function openReplicaSet(){ sed -i 's/#keyFile/keyFile/g' /etc/mongod.conf sed -i 's/#replication/replication/g' /etc/mongod.conf sed -i 's/#replSetName/replSetName/g' /etc/mongod.conf } function stopMongo(){ systemctl stop mongod } function startMongo(){ systemctl start mongod } function main(){ initenv inst initdb stopMongo cpReplicaSetKey openReplicaSet startMongo } main | tee -a ./log/inst_mongodb.log
4、安裝mongodb arbiter節點(在rs.addArb()之前其實還不能說是arbiter節點,只是起了個名字叫arbiter節點)

#!/bin/bash basedir=$(cd "$(dirname "$0")";pwd) function initenv(){ setenforce 0 sed -i 's/SELINUX=enforcing/SELINUX=disabled/g' /etc/selinux/config } ########inst mongdb 4.0############# function inst(){ cp -f ./mongdb/mongod-arbiter.conf /etc/ cp -f ./mongdb/mongod-arbiter.service /usr/lib/systemd/system/ mkdir -p /home/mongodb-arbiter/log mkdir -p /home/mongodb-arbiter/db chown -R mongod:mongod /home/mongodb-arbiter systemctl enable mongod-arbiter systemctl start mongod-arbiter } function initdb(){ cd $basedir cat ./mongdb/initUser.js | mongo --port 27018 --shell cat ./mongdb/initIndex.js | mongo --port 27018 -u "root" -p "abc123" --authenticationDatabase "admin" } function openReplicaSet(){ sed -i 's/#keyFile/keyFile/g' /etc/mongod-arbiter.conf sed -i 's/#replication/replication/g' /etc/mongod-arbiter.conf sed -i 's/#replSetName/replSetName/g' /etc/mongod-arbiter.conf } function stopMongo(){ systemctl stop mongod } function startMongo(){ systemctl start mongod } function main(){ initenv inst initdb stopMongo openReplicaSet startMongo } main | tee -a ./log/inst_mongodb_arbiter.log
5、初始replica set

#!/bin/bash basedir=$(cd "$(dirname "$0")";pwd) source ./app.properties nowtime=`date --date='0 days ago' "+%Y%m%d%H%M%S"` daytime=`date --date='0 days ago' "+%Y%m%d"` logPath=./log/init_repliaset_$daytime.log function log(){ echo $1 | tee -a $logPath } function logfile(){ echo $1 >> $logPath } function testConfig(){ log "check relate config." if [ "$masterServerIp" == "masterIp" ];then log "ERROR: masterServerIp is not correct in the file app.properties" exit 1 fi if [ "$slaveServerIp" == "slaveIp" ];then log "ERROR: slaveServerIp is not correct in the file app.properties" exit 1 fi local grepIp=`ifconfig | grep $masterServerIp` ## it is must execute in the master server which you input in the app.properties file when first initiate ths replica set if [ "$grepIp" == "" ];then log "the current server is not the master server, please login on to $masterServerIp and execute this script." exit 1 fi ## test the connection echo "use admin" > test.js echo "rs.slaveOk()" >> test.js echo "show dbs" >> test.js cat test.js | mongo --host $slaveServerIp -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27017 in $slaveServerIp server can not connect." exit 1; fi cat test.js | mongo --host $slaveServerIp --port 27018 -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27018 in $slaveServerIp server can not connect." exit 1; fi cat test.js | mongo --host $masterServerIp -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27017 in $slaveServerIp server can not connect." exit 1; fi rm -rf test.js log "check relate config end." } function initTwoMemberReplicaSet(){ log "init two member replica set." echo "use admin" > init.js echo "config = {_id:\"mongoSet\", members:[{_id:0,host:\"$masterServerIp:27017\",priority:2},{_id:1,host:\"$slaveServerIp:27017\",priority:1}]}" >> init.js echo "rs.initiate(config, {force:true})" >> init.js cat init.js | mongo -uroot -pabc123 --authenticationDatabase "admin" >> $logPath ## sleep 3 second for primary election. sleep 3s log "init two member replica set end." } function isElectPrimary(){ log "check is the master node selected." local masterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "PRIMARY" -B 4|grep "name" | grep $masterServerIp` local count=0 while [[ "$masterStr" == "" ]];do sleep 1s count=$((count+1)) masterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "PRIMARY" -B 4|grep "name" | grep $masterServerIp` if [ "$masterStr" != "" ];then log "the master node is selected." break fi if [ "$count" == "30" ];then log "ERROR: arbiter configure timeout, please configure manually with 'rs.addArb(\"$slaveServerIp:27018\")' in $masterServerIp:27017" exit 1 fi done log "check is the master node selected end." } function addArbiter(){ log "add arbiter ..." mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.addArb(\"$slaveServerIp:27018\")" log "add arbiter end." } function chekArbiter(){ log "check arbiter." local arbiterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "$slaveServerIp:27018" -A 3| grep "stateStr"` if [ "$arbiterStr" != "ARBITER" ];then log "ERROR: abriter $slaveServerIp:27018: $arbiterStr, may be you need restart mongodb with 'systemctl restart mongod-arbiter' command in $slaveServerIp." fi log "check arbiter end." } logfile "**************************$nowtime**********************************" testConfig initTwoMemberReplicaSet isElectPrimary addArbiter chekArbiter log "init end..."
6、重新配置replica set

#!/bin/bash basedir=$(cd "$(dirname "$0")";pwd) source ./app.properties nowtime=`date --date='0 days ago' "+%Y%m%d%H%M%S"` daytime=`date --date='0 days ago' "+%Y%m%d"` logPath=./log/init_repliaset_$daytime.log function log(){ echo $1 | tee -a $logPath } function logfile(){ echo $1 >> $logPath } function testConfig(){ log "check relate config." if [ "$masterServerIp" == "masterIp" ];then log "ERROR: masterServerIp is not correct in the file app.properties" exit 1 fi if [ "$slaveServerIp" == "slaveIp" ];then log "ERROR: slaveServerIp is not correct in the file app.properties" exit 1 fi ## get current replica set master node local currMasterNode=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "PRIMARY" -B 4|grep "name" | cut -d ":" -f 2 | sed 's/\"//g'` if [ "`ifconfig | grep $currMasterNode`" == "" ];then log "the current server has no mongodb master node, please login on to $currMasterNode execute this script." exit 1 fi ## test the connection echo "use admin" > test.js echo "rs.slaveOk()" >> test.js echo "show dbs" >> test.js cat test.js | mongo --host $slaveServerIp -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27017 in $slaveServerIp server can not connect." exit 1; fi cat test.js | mongo --host $masterServerIp -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27017 in $slaveServerIp server can not connect." exit 1; fi cat test.js | mongo --host $slaveServerIp --port 27018 -uroot -pabc123 --authenticationDatabase "admin" >> $logPath if [ $? != 0 ];then log "mongodb instance which port is 27018 in $slaveServerIp server can not connect." ## 有時,當主從服務器都重啟之后,有一個仲裁節點由於不在replica set中,但配置文件中卻存在replica的配置,造成不能夠正常啟動,但reconfig replica set后,需要將該節點reconfig到replica set中,故此處不能因為不能連接就執行exit,但輸出錯誤信息給出提示,即使服務未啟動, ## 執行rs.addArb()也可以添加仲裁節點,但必須確定添加的仲裁節點會啟動,否則即使通過rs.remove刪除掉了該仲裁節點,但主節點日志中也會一直打印錯誤信息, fi rm -rf test.js log "check relate config end." } function initTwoMemberReplicaSet(){ log "reconfig two member replica set." echo "use admin" > reconfig.js ## the _id must be the same as before ## find the master server id in previous replica set local masterId=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "$masterServerIp:27017" -B 1|grep "_id" | cut -d ":" -f 2|sed 's/,//g'` ## find the slave server id in previous replica set local slaveId=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "$slaveServerIp:27017" -B 1|grep "_id" | cut -d ":" -f 2|sed 's/,//g'` echo "config = {_id:\"mongoSet\", "protocolVersion" : 1, members:[{_id:$masterId,host:\"$masterServerIp:27017\",priority:2},{_id:$slaveId,host:\"$slaveServerIp:27017\",priority:1}]}" >> reconfig.js echo "rs.reconfig(config, {force:true})" >> reconfig.js cat reconfig.js | mongo -uroot -pabc123 --authenticationDatabase "admin" >> $logPath ## sleep 3 second for primary election. sleep 3s log "reconfig two member replica set end." } function isElectPrimary(){ log "check is the master node selected." local masterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "PRIMARY" -B 4|grep "name" | grep $masterServerIp` local count=0 while [[ "$masterStr" == "" ]];do sleep 1s count=$((count+1)) masterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "PRIMARY" -B 4|grep "name" | grep $masterServerIp` if [ "$masterStr" != "" ];then log "the master node is selected." break fi if [ "$count" == "30" ];then log "ERROR: arbiter configure timeout, please configure manually with 'rs.addArb(\"$slaveServerIp:27018\")' in $masterServerIp:27017" exit 1 fi done log "check is the master node selected end." } function addArbiter(){ log "remove arbiter in the master server." ## master server do not deploy arbiter, when reconfig, after the master node reselected you must specify the --post for delete or add local beforeArbiterIp=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "ARBITER" -B 4|grep "name" | cut -d ":" -f 2 | sed 's/\"//g'` if [ "`echo $masterServerIp | grep $beforeArbiterIp`" != "" ];then log "remove $masterServerIp:27018 in replica set." mongo -uroot -pabc123 --authenticationDatabase "admin" --host $masterServerIp --eval "rs.remove(\"$masterServerIp:27018\")" fi log "remove arbiter in the master server end." log "add arbiter ..." mongo -uroot -pabc123 --authenticationDatabase "admin" --host $masterServerIp --eval "rs.addArb(\"$slaveServerIp:27018\")" log "add arbiter end." } function chekArbiter(){ log "check arbiter." local arbiterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "$slaveServerIp:27018" -A 3 | grep "ARBITER"` ## arbiter connect need time, set to 15 seconds local count=0 while [[ "$arbiterStr" == "" ]];do sleep 1s count=$((count+1)) arbiterStr=`mongo -uroot -pabc123 --authenticationDatabase "admin" --eval "rs.status()" | grep "stateStr" -B 4|grep "$slaveServerIp:27018" -A 3 | grep "ARBITER"` if [ "$arbiterStr" != "" ];then log "the arbiter node is selected." break fi ## wait for 15 seconds if [ "$count" == "15" ];then log "ERROR: abriter $slaveServerIp:27018: $arbiterStr, may be you need restart mongodb with 'systemctl restart mongod-arbiter' command in $slaveServerIp." break fi done log "check arbiter end." } logfile "**************************$nowtime**********************************" testConfig initTwoMemberReplicaSet isElectPrimary addArbiter chekArbiter log "init end..."
腳本中通過rs.addArb()添加未啟動或者不存在的arbiter節點,可以添加成功,但主節點日志中會一直打印該arbiter節點相關的錯誤信息,即使通過rs.remove()刪除掉該arbiter節點,錯誤信息也會打印,該問題,重啟主節點可以解決
7、配置文件app.properties

masterServerIp=masterIp
slaveServerIp=slaveIp
執行初始化和重新配置replica set之前需要確認app.properties中配置是正確的
8、無論是初始化replica set或者重新配置replica set,前幾次都會出現not reachable/healthy,此時根據shell頁面打印的提示,在對應的服務器上執行systemctl restart mongod-arbiter即可