[Dump-users] Incremental backup features of dump

[Brent B. <br...@ca...>]

Hello, there are a few things about dump that I have always suspected but 
never have had the opportunity to test to see if they are true:


1)  Will dump track file deletions between incremental backups?  Here is
an example of what I mean by that:

Say I take a full backup of a directory full of files.  Sometime later, I 
delete some of the files -- deliberately (I don't want them back).  Then I 
do an incremental based on my last full backup (a level 1, for example).  
If I ever needed to recover from this backup set, due to something like a 
hard drive failure, would restore be smart enough to get rid of the files 
which disappeared between level 0 and level 1?  Or would it resurrect 
everything?

I've always presumed and *hoped* that it would delete such files during 
incremental restores.  I've based that hope on the notion that since dump 
and restore work with raw inodes and not just file contents like tar and 
cpio do, any incremental restore of a directory node that lost files since 
the last incremental would in effect "overwrite" that directory table with 
the newer copy that omits the deleted files, having the effect of 'rm' on 
the files that were pointed to.  Otherwise, if that were not true, you 
would get the following negative behavior during hard drive recovery:

Restore full backup (level 0):  Undesired files are extracted to disk
(naturally, because they were there at time of the full backup and hadn't
been deleted yet).

Restore incremental (level 1):  Undesired files are not in level 1, but 
they're left on the hard drive because you extracted them in level 0 and 
the incremental restore just leaves them there.  (After several such 
incremental restores, you could potentially resurrect a lot of cruft this 
way...)

Does dump avoid this problem?  I have always hoped that it would, since it 
would seem that one positive effect of extracting raw inodes is it would 
have the same effect as an 'rm' on any files that the inode pointed to 
before that were not in the newer incremental being extracted.


2)  If a file is copied onto a partition from another machine, and the
   file has a datestamp that's older than the latest dump recorded in
   /etc/dumpdates for that volume, will the next incremental catch the new
   file, on the premise that it didn't exist there before?  Or will it
   miss it because of the older datestamp?  This is actually a pretty
   common scenario, because one can only trust the datestamp to catch
   everything if one can assume that all files originate from the volume
   being backed up.  As soon as one starts bringing in files from
   elsewhere without changing their datestamps, files could get missed if 
   only the datestamp is used.


These are just a couple of points of curiosity...does anyone know?

-- 
+ Brent A. Busby, 	"The killer application for Windows was Visual
+ CATMIND RECORDINGS	 Basic. It allowed you to make your hokey, self-
+ br...@ca...	 made applications that did something stupid for
+ Pekin, IL   (USA)	 your enterprise...."  --Linus Torvalds