bug-tar
[Top][All Lists]
Advanced

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

[Bug-tar] using --no-check-device option with --listed-incremental + --o


From: Nathan Stratton Treadway
Subject: [Bug-tar] using --no-check-device option with --listed-incremental + --one-file-system
Date: Fri, 2 Sep 2011 16:08:26 -0400
User-agent: Mutt/1.5.20 (2009-06-14)

I was looking through the tar documentation and the incremen.c source
file, and wanted to make sure I correctly understand logic for looking
at the device/inode combo when running in --listed-incremental mode.

I gather that the basic idea is to try to notice when a directory that
was backed up in the previous incremental run has been moved out of the
way and replaced by a new directory with with same name (so that tar can
be sure to include the children of that new directory in the current
run's archive).

In looking over the code, it seems like the device id number is included
when comparing the previous and current directory info because only the
device id+inode combo is officially guaranteed to be unique in a given
system.  At the same time, it's clear from the way NFS filesystems are
handled that in practice doing just the inode portion of the check is
good enough in most situations.

I'm curious to know more about those situations where it wouldn't be
good enough...



Specifically, in my case I'm running tar in the context of Amanda, and
so my backups happen with both --listed-incremental and
--one-file-system... and thus all the directories being backed up in
each run have the same device id anyway.

So, my question is this: am I correct there is really no possible harm
from specifying --no-check-device on the tar invocation, too?  Or is
there some situation lurking out there where having --no-check-device
added on to --listed-incremental/--one-file-system would cause some sort
of problem?

(In my case, I haven't yet noticed any situations where my device ids
for particular filesystems have changed, but it seems like there are
many situations that cause such changes these days.  So I'm wondering if
there's any reason not to start using --no-check-device as a general
rule now, to make sure that we don't get bitten by a device-id switch at
some point in the future.)

Thanks.

                                                Nathan


----------------------------------------------------------------------------
Nathan Stratton Treadway  -  address@hidden  -  Mid-Atlantic region
Ray Ontko & Co.  -  Software consulting services  -   http://www.ontko.com/
 GPG Key: http://www.ontko.com/~nathanst/gpg_key.txt   ID: 1023D/ECFB6239
 Key fingerprint = 6AD8 485E 20B9 5C71 231C  0C32 15F3 ADCD ECFB 6239



reply via email to

[Prev in Thread] Current Thread [Next in Thread]