[BBLISA] dump or Legato Networker?

[John Stoffel]

>>>>> "Dean" == Dean Anderson <dean at av8.com> writes:

Dean> On Sun, 6 Jan 2008, Edward Ned Harvey wrote:
>> 
>> Is there any way to signal the Linux kernel to flush its filesystem write
>> cache, directly before performing the backup?

Dean> sync(8) should do this.  But maybe not on linux... sigh.

The problem is that this will only check point the filesystem(s) for a
few seconds at most.  If you have any processes which are still in the
process of writing to file(s), then they can still cause problems.

Supposedly there is support in ext3 so that when you take an LVM
snapshot, it will freeze the filesystem, (but not applications!) in a
self consistent manner so you can then mount the snapshot and back it
up, knowing that it's in good shape.

Again, you still need to make sure all your applications are in a
quiescent state as well.  In practice, it's too tough to do so, esp on
a general purpose system so people accept that there's some potential
to have corruption in a few files.


Also, as said before, there's the second problem of doing incremental
backups, because you need to go through once finding what changed,then
go back through again dumping the changes.  That leaves a window open
where a new file can be created but not backed up.  Again, it's
something that people just accept because it's very hard to do this
perfectly, though snapshots are helping.

Cheers,
John