21

I want to switch from MySQL to MongoDB but great data losses (more than 1 hour) are not acceptable for me.

I need to have 3 backup plans:

  1. Hourly backup plan. Data is flushed to disk every X minutes and if something wrong with the server I shall be sure that after reboot it will have all data at least for an hour ago. Can I configure it?

  2. Daily backup plan. Data is synced to backup disk every day so even if server explodes I can recover data for yesterday in some hours. Should I use fsync, master-slave or something else? I would like to have minimal traffic so ideally only changes will be sent.

  3. Weekly backup plan. Data is synced to second backup disk so if both server and first backup disk explode I have at least data for last week. Here this is the question of reliability so it's ok to send all data via network.

How can I do it?

luchaninov
  • 6,792
  • 6
  • 60
  • 75

5 Answers5

21

  1. The fsync command flushes the data to disk. It is executed each 60 seconds by default, but can be configured using the --syncdelay command line parameter.

  2. The documentation on backups has some good pointers for daily and weekly backups. For the daily backup, a master-slave configuration seems like the best option, as it will only sync changes.

  3. For the weekly backup you can also use a master-slave configuration, or replication. Another option is the mongodump utility, which will back-up the entire database. It is capable of creating backups while the database is running, so you can run it on the main database or one of the slaves. You can also lock the slave before backing it up.

Niels van der Rest
  • 31,664
  • 16
  • 80
  • 86
0

If you want to outsource the backup solution entirely, MongoDB Management Service takes snapshots every six hours. The default retention policy on the snapshots will allow you to get point-in-time restore for 24 hours, daily snapshots for a week, weekly snapshots for a month, and monthly snapshots for a year.

This FAQ has the full retention policy.

The backup service continually backs up your replica set by reading the oplog so the overhead is lower than full local periodic snapshots.

user780232
  • 3
  • 1
erlichson
  • 174
  • 5
0

May be you can use automongobackup .

wanghao
  • 3,335
  • 2
  • 18
  • 13
0

Try this backup script if you want to create a backup from slave mongodb database to S3.

  1. DB host (secondary preferred as to avoid impacting primary performance)

    HOST='SomeHost/mongodbtest-slave'

  2. DB name DBNAME=***

  3. S3 bucket name BUCKET=*-backup

  4. Linux user account USER=ubuntu

  5. Current time TIME=/bin/date +%d-%m-%Y-%T

  6. Password PASSWORD=somePassword#!2*1

  7. Username USERNAME=someUsername

  8. Backup directory DEST=/home/ubuntu/tmp

  9. Tar file of backup directory TAR=$DEST/../$TIME.tar

  10. Create backup dir (-p to avoid warning if already exists) /bin/mkdir -p $DEST

  11. Log echo "Backing up $HOST/$DBNAME to s3://$BUCKET/ on $TIME";

  12. Dump from mongodb host into backup directory

mongodump --port 27017 -d DBNAME -u USERNAME -p PASSWORD -o $DEST

  1. Create tar of backup directory /bin/tar cvf $TAR -C $DEST .

  2. Upload tar to s3 /usr/bin/aws s3 cp $TAR s3://$BUCKET/

  3. Remove tar file locally /bin/rm -f $TAR

  4. Remove backup directory /bin/rm -rf $DEST

All done echo "Backup available at https://s3.amazonaws.com/$BUCKET/$TIME.tar

You can use the steps above put them in a shell executable file and execute this at any interval using crontab commands.

Pragyanshu Sharma
  • 374
  • 3
  • 12
0

On the first point.

MongoDB has a term like 'durable write operation'. If journaling is enabled, you may only lose data that has not been writed in the log. This is a very small amount of time (100 milliseconds by default)

On the second and third points.

You can set master slave replication, but this will not protect you from data errors (for example, if important data is accidentally deleted). Therefore, you need to provide regular backups one way or another.

There are several approaches here:

  1. LVM snapshot is a good solution if your filesystem supports it and the journaling is enabled in your mongoDB. For more details see here
  2. mongodump -You can create a shell script that will run scheduled backups via cron and send them to storage. Here is an example of a good script
  3. Backup as a Service. There are many solutions to make the backup for you.
Ivan
  • 241
  • 1
  • 10