3

I am attempting to set up an automatic backup for a server running Ubuntu. This server is used to host a couple applications which are rather write heavy and subject to a good amount of traffic. To be specific, a Solr and a MySQL database is being maintained and is constantly updated/written to. The amount of data on the server is currently in the range of 20GB, but within a year it should grow to be at least 5TB. So there isn't a lot of data on the server right now, but the solution I come up with should be scalable in the future.

I want to set up a backup of the whole server including the Solr & MySQL indexes, but I am unsure about how to go about it. So far I've set up Tartarus to do an incremental backup, however I've realised that it will be problematic to perform it with only this tool as the server doesn't have LVM enabled. Since LVM isn't enabled I'm getting corrupt/inconsistent backups as files are being written to while being backed up. Is it a correct assumption that I want the ability to make a LVM snapshot of the system and then perform a backup?

I've also entertained the idea that the backup should be split into several parts. One part responsible for handling Solr, one handling MySQL, and one handling the rest of the system. Is this a viable option?

-- UPDATE--

I ended up opting for the approach where I split the backup. I do the Solr backup using replication and then ftp to the backup server. For MySQL I use a mysqldump for now and ftp that. For the rest of the data I use Tartarus with an incremental backup.

Roman
  • 3,907
  • 3
  • 21
  • 34
PetaPetaPeta
  • 33
  • 4

1 Answers1

1

As far as MySQL is concerned the preferred backup method is to setup a replica slave and then take the backup there.

See, e.g.:

http://dev.mysql.com/doc/mysql-backup-excerpt/5.6/en/replication-solutions-backups.html

As for solr, perhaps an incremental rsync solution might work? It seems to depend on the version.

Community
  • 1
dmourati
  • 25,540
  • 2
  • 42
  • 72
  • So the preferable way of performing this task is to split the backup into several disjoint parts? One for Solr, one for MySQL, and one for the rest. – PetaPetaPeta Jul 03 '13 at 08:27
  • 2
    Database backups are always subject to special requirements so, yes, a per-service style backup seems appropriate when considering MySQL, solr, plus base OS. – dmourati Jul 03 '13 at 08:30
  • So would you say that it is neccessary to set up LVM? – PetaPetaPeta Jul 03 '13 at 08:49
  • 1
    LVM is nice to have but I wouldn't call it necessary under most circumstances. Note well, many Linux OS distributions default to setting up LVM during install. – dmourati Jul 03 '13 at 08:58
  • Okay, thanks a lot for your help. I will update my original question with the final decisions to this when I figure out exactly what to do. – PetaPetaPeta Jul 03 '13 at 09:09