Push vs Pull Backups. What do you prefer?

nfn

Hi,

I'm revisiting my backup strategy and would like to ask you what are your preferences when dealing with multiple VPS's?
Do you prefer Push or Pull backups? Which tools would you use for each one?

Thanks

bnmkl

Pull.

sftp for security, or wget/curl for speed.

rsync for sync.

Corey

@nfn said: Do you prefer Push or Pull backups?

I guess I do not understand what you are asking. Pushing backups to a remote location or remote location pulling from you? Either way you are pushing and pulling.

Abdussamad

Pull is safer and it is what I use. With push you will loose both files on both the production machine and the backup machine if the production box were to be compromised. Since the production box is running a lot of services it is more likely to be exploited compared to the backup box that will do just backups.

ShardHostSarah

We push then pull.

bnmkl

@ShardHost ... We push then pull.

Hawt !

nfn

Anyone using rdiff-backup for pull?

twain

@Abdussamad said: Since the production box is running a lot of services it is more likely to be exploited compared to the backup box that will do just backups.

good point, for this reason, I'd generally agree that pulling is safer security-wise. backuppc (rsync pull from the backuppc host) ftw

netomx

Am I getting old, or is this a deja vú? I remember this question

raindog308

@Abdussamad said: With push you will loose both files on both the production machine and the backup machine if the production box were to be compromised.

This.

twain

@raindog308 said: @Abdussamad said: With push you will loose both files on both the production machine and the backup machine if the production box were to be compromised.

This.

However, with pull, if your backup host get compromised, say a possible bye-bye to all of your production machines (that your backup host pulls from)

jcaleb

pull

Abdussamad

@twain said:

However, with pull, if your backup host get compromised, say a possible bye-bye to all of your production machines (that your backup host pulls from)

That is why I wrote:

@Abdussamad said: Since the production box is running a lot of services it is more likely to be exploited compared to the backup box that will do just backups.

You're taking a risk with either push or pull but the risk is less with pull. Also if you can afford multiple production servers you can also afford multiple backup VPS - one for each production server. That way any exploit in a single backup VPS will only affect one production server.

natestamm

@nfn i would say at the risk of being obstinate but ya'll know that doesn't bother me...Pull + rsync for the win
If it helps modern trend as any University will tell you in Sales and SCM is all based on Pull
..seems community agrees also!

pubcrawler

I don't see any real benefit of push or pull as posed.

Typically we pull from a main production node via a remote backup node. Helps to isolate records of the backups. Either way, has its hazards and needs to have backup reports/sanity checks.

raindog308

@twain - uh, how does that work?

Backups pulls via rsync (real rsync, not over ssh). If the backup server is compromised, best the attacker can do is nuke the backups.

nfn

So, what options do we have for pull backups that don't need direct file access? Somethig like Bacula but easier and light.

Any recomendations?

Oliver

Push is useless from a security perspective.

Pull for everything.

sleddog

Push from production box to backup box via rsync over ssh.
Backup box uses authprogs to restrict what the user can do (i.e., only run the backup script).
On the backup box, run rdiff-backup against the synced data, creating redundancy in a separate directory structure (inaccessible from the production box).
Pull the rediff'ed data (rsync/ssh) to a second backup box for further redundancy.