Here's my thoughts/rant on backup systems. First of all, you need to
consider which problems you're trying to solve with backups:
* Hardware failure - To deal with simple hardware failure like the
drive itself failing, you just need a copy of the data. The amount
of time between copies is the amount of risk you're exposing
yourself to. If you can deal with redoing a days' work, then
nightly backups/copies are good enough. RAID 1/mirrored drives
provide an instantaneous copy. Rsync fired via cron can provide
other windows of risk, like hourly, etc. Pick the amount of risk
you're willing to deal with.
* Human failure - "Oh drat, I really do need that file I deleted last
week!" or "That edit turned out wrong, I wish I could go back to the
version I had 24hrs ago." To deal with this kind of problem, you
need multiple copies of your data spaced out over time. This is
where incremental backups help because keeping 30 copies of all your
data gets expensive real fast.
* Change tracking or Pointing the finger o' blame - Some systems have
a need or even a legal responsibility to track/restrict who can make
changes and to log when those changes were made. Source code control
like SVN/git work well here and can provide a complete history of
changes stretching back many years. But you're not going to put
commonly changed files like your email Inbox into git.
* Catastrophic failure - Things like fire, theft and police raids on
the data center seizing all the servers, or the data center going
out of business suddenly. This is where you need to have a copy of
your data in a separate location. This is also where encryption of
your backups becomes important. How much do you trust the other
location where a copy of your data resides?
* How much data are you dealing with here? - This is incremental vs
full backups. A writer can keep backups of their books in progress
on a flash drive tucked into their pocket. It'll take them 30
seconds to copy it back and forth. A musician or video editor is
going to have multiple TB of data to move around and that takes
significant time. The backup need to finish before the next work
period starts. If your office/business works multiple shifts, that
further reduces the window you have to do the backup.
My personal solution, which isn't going to fit everybody, is:
* SVN on a central file server for all my projects that require
long-term tracking.
* Nightly rsync of all my external server data to a large drive on a
server here in the office.
* For the couple of MySQL servers where I can't afford the nightly 15
minute down time to lock and then back up all of the data, I run a
replication server and lock that up instead. I keep 7 nightly
copies of the database dumps which are further backed up by the file
system backups.
* Retrospect (https://www.retrospect.com/) to backup all of the
machines here in the office. It's a windows product, but pretty
decent and it'll encrypt the backups. I do incremental backups
during the week to an external hard drive. On Friday, I swap the
drive and it does a full backup on Fri. nightly which takes ~16hrs.
I bring the other drive home and I've got 4 of them in rotation, so
I've got 4 weeks worth of copies.
My office is ~1 mile from my house. My usual joke is that if there's a
catastrophe big enough to take out both my office and my home, it's
probably taken out me as well, so at that point, I won't give a shit
about the backups being unrecoverable.
I hope some of that helps,
Rene
On 10/6/2016 10:08 PM, Joe Golden wrote:
I feel like I should get a little more serious about a backup system
and encryption on my filesystems.
I use ssh for command line access to texty things for projects, etc.
I use sshfs for mounting a shared directory for working space between
a server, desktop and a couple of laptops. The important bits live on
the server.
What's the standard recommended encryption for an encrypted home dir?
Looks like encfs is a good bet.
And for backups, Anthony mentioned Git Annex. Is that a backup
solution or something similar? In general I don't need incremental
backups, but if didn't cost much and made things faster, all the
better. I love git and think git should be in more places, and love
the distributed idea.
Sorry I know this is a big question. Any recommendations from the
list appreciated.
PS: we should do beerz sometime.
--
------------------------------------------------------------------------
René Churchill
VP of Development (i.e. Geek #2)
WherezIt.com - Your source for Local information
r...@wherezit.com <mailto:r...@wherezit.com>
802-244-7880 x527
http://www.wherezit.com/
