0

I'm looking for a backup strategy for my production server. It should be fully automatic and as "real-time" as possible.

Further, i don't want to do much to recover everything. Let's say like a backup server which we can switch to if the production server goes down.

I'm using the following services with Debian Lenny:

  • Apache2
  • MySQL 5.1
  • Postfix
  • Dovecot
  • SpamAssassin
  • SSH of course

So, i think i'll need a backup-server which replicates all files in near-real-time. Is this a well-known scenario? Can you tell me what i have to do to achieve this?

Further, i think i'll need Full- and incremental Backups of the System (each week Full and every 12 Hours incremental). How can i achieve this ?

This all should be easy and fast to recover ... Any help would be greatly appreciated !!

I don't use Plesk.

Stefan
  • 146
  • 7
  • Please complete your question with at least the amount of data you need to back up. It's a completely different thing to backup only small amounts of rarely changing data instead of terabytes of consistently changing data. – Janne Pikkarainen Sep 14 '11 at 08:41

1 Answers1

1

There's lots of questions, many of which already have answers here on Serverfault.

One strategy will not meet the requirements for all the software.

i think i'll need Full- and incremental Backups....should be easy and fast to recover

then forget about incremental backups.

For mysql, the best solution is to maintain a replicated slave which you can stop and run mysqldump against.

For the other stuff, the best solution with minimal downtime would be to have a mirrored filesystem - then when you want to backup,

  • stop the mail services,
  • flush the buffers,
  • break the mirror,
  • start up mail services
  • mount the mirror elsewhere
  • backup from the mirror
  • unmount the mirror
  • join it back into the volume group

Further, i don't want to do much to recover everything. Let's say like a backup server which we can switch to if the production server goes down

Not sure what you mean here. If you only want to provide for hardware failures/onsite recovery and you've already got a backup machine available, then run the mysql slave there, and put the data for the other applications on a shared drive - e.g. a volume maintained via DRBD, preferably with a cluter file system running on it - like GFS or Lustre - or use a replicated filesystem like AFS.

Community
  • 1
symcbean
  • 19,931
  • 1
  • 29
  • 49
  • +1 except I would not "forget about incremental backups". Smart use of incremental backups (e.g. Weekly full, Daily incremental ; Monthly full, Weekly differential, Daily incremental) can still be recovered without too much effort/time and saves a lot of space on the backup media. – voretaq7 Sep 14 '11 at 16:46
  • i found another solution that - i think - fits my needs better: using linux-ha with heartbeat, openais and pacemaker. thanks for your answer ! – Stefan Sep 15 '11 at 12:36