Like many of you I am "the guy who takes care of backup" in the company. Luckily we are a software shop and most of the mission critical data are kept in a bunch of mercurial repositories. This makes it extremely easy for backup - all you need a script to update a remote copies and clone new repos.
There are things that aren't simply version controlled however - like a large parts of our server(s) configuration, email, web data etc. For that we use a combination of rsnapshot and plain rsync to a zfs snapshotted box. No matter what we use - the end result is a bunch of files on a filesystem.
Given the amount of data these can't simply be copied over the internet to our off-site backup store (which is simply a white box sitting at my home) I am toying with the idea of moving hard drives between the office and my home as an insurance against corruption to the on-site copies - whatever the cause will be.
Doing that is also very simple - for our rsnapshot backup all I need to do is to clone the MD devices which the filesystem resides on while for the backup on ZFS we can simple do a zfs send and recv on a new disk/zpool..
Does the Sneakernet approach make sense for us? It's also useful to quickly explain the whole idea to my fellow workmates. :)
p.s. English is not my first language so don't be alarmed if you see broken grammar above :)
Answer
Ultimately, after all the various electronic methods, most major corporations have their backup tapes/disks taken by hand off-site for long-term storage. I myself rotate a bunch of firewire-connected hard disks that I backup my various machines to into and out of a safety deposit box every few weeks. So yes, 'sneakernet' is a good way of dealing with backup in my opinion.
No comments:
Post a Comment