背景：
最近后端基于mongo的项目越来越多，MySQL基于冷备份+binlog可以恢复至任意时间点，那么mongo是否有同样的功能呢？经过调研发现可以通过dump+oplog可以实现粒度更细致的恢复。
以下是实验步骤，附脚本内容。
1、 制作测试数据并且获取冷备
2、模拟线上两个冷备份之间的数据写入
3、备份oplog
4、冷备+oplog恢复至临时实例，确认没问题后，通过临时实例恢复到主库
1、
dump比较重要的几个参数
#mongodump --help
Usage:
mongodump
Export the content of a running server into .bson files.
Specify a database with -d and a collection with -c to only dump that database or collection.
See http://docs.mongodb.org/manual/reference/program/mongodump/ for more information.
connection options:
-h, --host= mongodb host to connect to (setname/host1,host2 for replica sets)
--port= server port (can also use --host hostname:port)
authentication options:
-u, --username= username for authentication
-p, --password= password for authentication
--authenticationDatabase= database that holds the user's credentials
--authenticationMechanism= authentication mechanism to use
--gzip compress archive our collection output with Gzip
--oplog use oplog for taking a point-in-time snapshot
test_1234:PRIMARY> show dbs
admin 0.000GB
local 2.724GB
test_1234:PRIMARY> use test
switched to db test
test_1234:PRIMARY> for (var i=0;i<=5000;i++)(db.test.insert({"test":i}))
WriteResult({ "nInserted" : 1 })
test_1234:PRIMARY> db.test.count()
5001
当前test.test有5001条数据，获取冷备份并查看当前时间戳
#date +%s
1553499626
2、 这个时候模拟下冷备之后的数据写入，删除一些数据并记录oplog
test_1234:PRIMARY> use test
switched to db test
test_1234:PRIMARY>
test_1234:PRIMARY>
test_1234:PRIMARY> show tables
test
test_1234:PRIMARY> for (var i=0;i<=5000;i++)(db.after.insert({"after":i}))
WriteResult({ "nInserted" : 1 })
test_1234:PRIMARY> db.test.remove({"test":{$gt:1500}})
WriteResult({ "nRemoved" : 3500 })
test_1234:PRIMARY> db.test.count()
1501
test_1234:PRIMARY> db.after.count()
5001
test_1234:PRIMARY> #date +%s
1553499686
3、备份oplog
./mongoha_dump_oplog.sh
options:
-S [Start-BSON-Timestamp]
-E [End-BSON-Timestamp]
is running like mongoha_dump_oplog.sh -S -E
#
./mongoha_dump_oplog.sh -S 1553499626 -E 1553499686
2019-03-25T15:42:07.579+0800 writing local.oplog.rs to
2019-03-25T15:42:10.576+0800 local.oplog.rs 0
2019-03-25T15:42:13.575+0800 local.oplog.rs 0
2019-03-25T15:42:16.575+0800 local.oplog.rs 0
2019-03-25T15:42:19.575+0800 local.oplog.rs 0
2019-03-25T15:42:22.575+0800 local.oplog.rs 0
2019-03-25T15:42:25.576+0800 local.oplog.rs 0
2019-03-25T15:42:26.267+0800 local.oplog.rs 8507
2019-03-25T15:42:26.267+0800 done dumping local.oplog.rs (8507 documents)
Mongo oplog backup [2019-03-25 15:40:26 - 2019-03-25 15:41:26] -->Sucess
4、 新建临时实例利用刚才冷备+oplog恢复，确认数据没问题之后，恢复至主库。
1）新建临时端口
mongo_install.sh -T 1 -P 6666 -M 1
2019-03-25 15:32:57 start install
mongodb-linux-x86_64-rhel62-3.4.0/README
mongodb-linux-x86_64-rhel62-3.4.0/THIRD-PARTY-NOTICES
mongodb-linux-x86_64-rhel62-3.4.0/MPL-2
mongodb-linux-x86_64-rhel62-3.4.0/GNU-AGPL-3.0
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongodump
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongorestore
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongoexport
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongoimport
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongostat
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongotop
mongodb-linux-x86_64-rhel62-3.4.0/bin/bsondump
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongofiles
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongooplog
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongoreplay
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongoperf
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongod
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongos
mongodb-linux-x86_64-rhel62-3.4.0/bin/mongo
/usr/local/mongodb is exists ..
mongo data,log dir create sucess
2019-03-25 15:32:57 ----> mongodb6666 start sucess
note: noprealloc may hurt performance in many applications
about to fork child process, waiting until server is ready for connections.
forked process: 8394
child process started successfully, parent exiting
2）恢复冷备+oplogmongoha_fullrestore.sh
options:
-P specify the port
-p specify the restore db path ,must abspath
is running like mongoha_backup.sh -P 27017 -p /test
mongoha_fullrestore.sh -P 6666 -p /data1/backup/mongodb/2019/03/25/_2019-03-25_15\:15/
at 2019-03-25_15:34 to do restore for mongo
2019-03-25T15:34:48.505+0800 preparing collections to restore from
2019-03-25T15:34:48.506+0800 reading metadata for test.test from /data1/backup/mongodb/2019/03/25/_2019-03-25_15:15/test/test.metadata.json
2019-03-25T15:34:48.544+0800 restoring test.test from /data1/backup/mongodb/2019/03/25/_2019-03-25_15:15/test/test.bson
2019-03-25T15:34:48.573+0800 restoring indexes for collection test.test from metadata
2019-03-25T15:34:48.573+0800 finished restoring test.test (5001 documents)
2019-03-25T15:34:48.573+0800 replaying oplog
2019-03-25T15:34:48.573+0800 done
Mongo restore() -->Sucess
冷备恢复成功
临时节点查看下数据
恢复到冷备之前的数据了，但是之前删除的数据及后插入的after document还未恢复，这时候我们恢复oplog
./mongoha_restore_oplog.sh
options:
-P specify the port
-p specify the restore db path ,must abspath
is running like ./mongoha_restore_oplog.sh -P 临时端口 -p /data1/backup/mongodb/oplog/2019/03/25/local/oplog.rs.bson
再次查看临时库数据
临时库数据以恢复，这时候就可以备份出来恢复至线上了。
附mongoha_fullbackup.sh脚本
#!/bin/bash
#auth [email protected]
dat=`date "+%Y-%m-%d_%H:%M"`
day=`date "+%Y/%m/%d" -d now`
#localip=`ip a | grep "\binet\b" |awk '{print $2}'| egrep -e "^10|172" | awk -F"/" '{print $1; exit;}'`
basedir="/data1/backup/mongodb/$day"
helpfunc(){
echo
echo "options:"
echo " -P specify the port"
echo " is running like mongoha_backup.sh -P 27017 -D test,test1"
}
while getopts "P:" Option
do
case $Option in
P) ports=$OPTARG;;
*) helpfunc; exit 1; ;;
esac
done
dbname=`echo $dbname | sed 's/,/ /g'`
dobackup(){
echo "at $dat to do bakcup for mongo$i"
mkdir -p "$basedir"
#mongodump -u$bacuser -p$pswd --port $i --oplog -o "$basedir"/mongo"$i"
if /usr/local/mongodb/bin/mongodump -h 127.0.0.1 --port $ports --oplog -o "$basedir"/"$i"_$dat;then
echo "Mongo Backup($i) -->Sucess"
else
echo "Mongo Backup($i) -->faild"
fi
}
check(){
person=`whoami`
if [ "$person" != "root" ]; then
echo "the user is now $person, checkout to 'root' to continue"
exit 1
fi
if [ -z "$ports" ] ;then
helpfunc
exit 1
fi
}
cleanbackup(){
rmdate=`date "+%Y/%m/%d" -d "7 days ago"`
if [ -d $1 ] ;then
rm -rf $1
echo "rm -$1"
elif [ -d $basedir/$rmdate ] ;then
rm -rf $basedir/$rmdate
echo "rm -rf $basedir/$rmdate"
else
rm -rf $basedir/mongo$1
echo "rm -rf $basedir/mongo$1"
exit 1
fi
}
main(){
check
dobackup
}
main
附mongoha_fullrestore.sh脚本
mongoha_fullrestore.sh
#!/bin/bash
dat=`date "+%Y-%m-%d_%H:%M"`
day=`date "+%Y/%m/%d" -d now`
#localip=`ip a | grep "\binet\b" |awk '{print $2}'| egrep -e "^10|172" | awk -F"/" '{print $1; exit;}'`
basedir="/data1/backup/mongodb/$day"
helpfunc(){
echo
echo "options:"
echo " -P specify the port"
echo " -p specify the restore db path ,must abspath"
echo " is running like mongoha_backup.sh -P 27017 -p $backdir/test"
}
while getopts "P:p:" Option
do
case $Option in
P) ports=$OPTARG;;
p) path=$OPTARG;;
*) helpfunc; exit 1; ;;
esac
done
#mongorestore -h 127.0.0.1:27017 -d test test/
dorestore(){
echo "at $dat to do restore for mongo${dbname}"
if /usr/local/mongodb/bin/mongorestore --drop --oplogReplay -h 127.0.0.1:${ports} $path;then
echo "Mongo restore(${dbname}) -->Sucess"
else
echo "Mongo restore(${dbname}) -->faild"
fi
}
check(){
person=`whoami`
if [ "$person" != "root" ]; then
echo "the user is now $person, checkout to 'root' to continue"
exit 1
fi
if [ -z "$ports" ] ;then
helpfunc
exit 1
fi
}
main(){
check
dorestore
}
main
附mongoha_dump_oplog.sh 脚本
cat mongoha_dump_oplog.sh
#!/bin/bash
dat=`date "+%Y-%m-%d_%H:%M"`
day=`date "+%Y/%m/%d" -d now`
helpfunc(){
echo
echo "options:"
echo " -S [Start-BSON-Timestamp]"
echo " -E [End-BSON-Timestamp] "
echo " is running like mongoha_dump_oplog.sh -S -E "
}
if [ $# -lt 2 ] ;
then
helpfunc
exit 1
else
while getopts "S:E:" Option
do
case $Option in
S) start_t=$OPTARG;;
E) stop_t=$OPTARG;;
*) helpfunc; exit 1; ;;
esac
done
fi
start_time="Timestamp(${start_t}, 0)"
stop_time="Timestamp(${stop_t}, 0)"
date_now=`date "+%Y%m%d%H_%M" -d now`
backdir="/data1/backup/mongodb/oplog"
#dump
function data_fetch(){
mkdir -p $backdir
query='{ts:{$gt:'$start_time',$lt: '$stop_time' }}'
flags='-d local -c oplog.rs'
if /usr/local/mongodb/bin/mongodump $flags --query "$query" -o $backdir/$date_now;then
echo "Mongo oplog backup [`date -d @${start_t} "+%Y-%m-%d %H:%M:%S"` - `date -d @${stop_t} "+%Y-%m-%d %H:%M:%S"`] -->Sucess"
else
echo "Mongo oplog backup [`date -d @${start_t} "+%Y-%m-%d %H:%M:%S"` - `date -d @${stop_t} "+%Y-%m-%d %H:%M:%S"`] -->Faild"
fi
}
main(){
data_fetch
}
main
附mongoha_restore_oplog.sh 脚本
cat mongoha_restore_oplog.sh
#!/bin/bash
dat=`date "+%Y-%m-%d_%H:%M"`
day=`date "+%Y/%m/%d" -d now`
#localip=`ip a | grep "\binet\b" |awk '{print $2}'| egrep -e "^10|172" | awk -F"/" '{print $1; exit;}'`
backdir="/data1/backup/mongodb/oplog/$day"
helpfunc(){
echo
echo "options:"
echo " -P specify the port"
echo " -p specify the restore db path ,must abspath"
echo " is running like $0 -P 临时端口 -p $backdir/local/oplog.rs.bson"
}
if [ $# -lt 2 ] ;
then
helpfunc
exit 1
else
while getopts "P:p:" Option
do
case $Option in
P) port=$OPTARG;;
p) path=$OPTARG;;
*) helpfunc; exit 1; ;;
esac
done
fi
function data_restore(){
flags='-d local -c oplog.rs'
if /usr/local/mongodb/bin/mongorestore --port $port $flags $path;then
echo "Mongo restore[$port] -->Sucess"
else
echo "Mongo restore[$port] -->Faild"
fi
}
data_restore