[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Live Fille System Backup

Hash: SHA1

On Mon, May 08, 2017 at 04:03:54PM -0700, Sergei G wrote:
> And if you think application consistency is nothing to worry about know
> that Git on Ext4 has failed to recover.
> My current conclusion:
> 1. If I want application consistency I have to take down any application
> that has an internal database or backup those applications prior to FS
> backup.

That depends. There are applications designed with (on-disk) consistency
in mind (I mention PostgreSQL, I start sounding like a broken record).

But whether you have to "take down" the application or just "alert" it
for the time the snapshot takes, without its cooperation you either risk
inconsistent state or bad performance (there are cases where apps chose
number two, because they aren't designed for high throughput anyway).

The gist is:

  (a) file system magic can't "do" it alone
  (b) most of the time, a naive "snapshot" with rsync will do
  (b') you alone can decide, based on your apps & requirements,
      whether the above "most of the time" works for you

With snapshotting file systems (be it native or assisted) you get just
faster snapshots, i.e. you mitigate the problem. If you want to go all
the way, you have to get the app's cooperation, either voluntary (it's
designed for that) or forced (shutdown)

There's no magic.

- -- tomás
   "all generalizations suck"
Version: GnuPG v1.4.12 (GNU/Linux)


Reply to: