Three basic backup methods of mongodb

Catalog

One 、 snapshot

Two 、 Data file copy

3、 ... and 、mongodump And mongorestore

---

One 、 snapshot

        Using snapshots for backup is the most common method , Many enterprise backup products are based on this （ Love number 、commvault etc. ）. Take the backup schematic diagram of love digital as an example .

1. The management console starts the backup task .
2. The console sends backup commands to each physical client , Each client finds whether the local node is Primary（ Lord ） node , If not Primary（ Lord ） node , The backup task of this node ends .
3. Primary（ Lord ） The client of the node gets MongoDB The file system where the data directory is located , Create a snapshot volume for the file system and mount it to a temporary directory .
4. The client transmits the data of the temporary directory to the storage server .
5. Client completes backup , Report the task backup status to the management console , End the entire backup task .

  The main advantage of this method is its high speed , Easy to use .

The disadvantages are obvious ：

1、 All hosts in the replica set must be preserved   > 500MB Of VG Volume space . Although only the primary node performs backup operations , However, in replica set architecture, all nodes may become master nodes .

2、 Unable to achieve incremental backup , The reason is simple , It cannot determine the change of data , It is not clear what changes have taken place between the data at this point in time and the data at the last backup time .

3、 Only right mongod All databases under the instance are collectively backed up , Unable to achieve fine-grained database backup .

4、 Snapshot data recovery is also quite tedious , Especially remote recovery .

Two 、 Data file copy

        This method is also relatively simple and direct , Direct will mongoDB Copy all the data under the data directory . But there are a few things to note .

1、 Before copying, you need to lock the database . You cannot write to the database after locking , May affect normal business . Remember to unlock after copying .

2、 You need to stop when you recover mongod example .

        The disadvantages of this approach are similar to snapshots . Specific operation , As shown below ：

#  Get into mongo shell
db.fsyncLock()

#  Return to linux shell
#  take mongo Of dbpath Copy all to the cache directory 
cp -R /var/lib/mongo/* /backup

#  Get into mongo shell
db.fsyncUnlock()

3、 ... and 、mongodump And mongorestore

        These are both mongo Integrated backup and recovery tools , The function is also very powerful .

        Use mongodump Database data can be exported as binary bson File format . And support collection Granular backup .

        mongorestore It's cooperation mongodump The use of , As long as it is mongodump All exported data can be replied .

        And cooperate with mongoDB Enabled in replica set mode oplog function , Incremental backup can be achieved .

        Their only drawback is that their performance is too low , Not as good as the above two methods .github It has go Language implementation of the source code , If you have enough energy, you can read and optimize .

        Because of the length （ lazy ）, See the following link for specific usage mongodump — MongoDB Manualhttps://www.mongodb.com/docs/v4.4/reference/program/mongodump/
mongorestore — MongoDB Manualhttps://www.mongodb.com/docs/v4.4/reference/program/mongorestore/