denis cardon <cardondenis < at > yahoo.com>
wrote the following on Thu, 8 Jan 2004 05:48:52 -0800 (PST)
However, since the internet connection is vital for running the
business (vpn stuff et al.), the backup shouldn't be lasting after
8:00AM. So if there has been a lot more than 500MB of data change on
the server I need to kill the backup process.
It is not a huge problem if the file backup is late one
day. However, since rdiff does regression, if one day backup can't
get finished, it will never get finished again (always more data to
backup since last snapshot).
So my question is : is there a way to disable regression? If not, is
there a fundamental flaw in having a partial snapshot of the
file-server (I backup only documents, no /etc/, no /usr/, no /var/).
Sorry, there's no way to disable regression. It's a pretty hard
problem to figure out a way not to lose data and keep the archive
consistent no matter what is happening to the source side or how/when
rdiff-backup gets killed. Regression is the best I've come up with so
far, although I have to admit it's slow, so it's a pain when a backup
session fails.
Anyway, in your case, instead of stopping rdiff-backup, and starting a
new session the next day, could you just pause the rdiff-backup
session, and let it finish the next night? Then there would be no
regression because no session actually failed.
I'm not sure how to pause processes; it probably depends on the
operating system. One cheap way would be to run rdiff-backup in a
window and then press Control-S to pause, Control-Q to resume. Of
course this way doesn't seem to lend itself to automation...
--
Ben Escoto
