[rdiff-backup-users] failing backup

[Kurt Yoder]

Hello list

I am running rdiff-backup 1.1.5 on a Debian machine. Previously I  
have been able to do restores without problems. However, now I am  
having trouble restoring my boss's mail directory. It's his "in-box"  
maildir directory (eg "./cur") with 2743 files in it, the largest  
being 1.1 MB. Whenever I try to restore this file using -v9, I see  
the preparation activity when rdiff-backup launches.

Then rdiff-backup does... *something* which is not displayed on the  
terminal. The rdiff-backup process starts to use up progressively  
more memory, until after several hours it is using gigabytes of  
memory. At this point it segfaults and nothing in the maildir is  
restored.

I did a bit of further digging with strace, and see an un-ending  
stream of these kinds of messages:

lstat64("/backup/disk1/ns2_maildir/rdiff-backup-data/increments/ 
richard-group.com/david/cur/ 
1170338212.M12781P21813V0000000000000801I03130440_0.db1,S=3073:2,S. 
2007-02-08T05:18:56-05:00.snapshot.gz", {st_mode=S_IFREG|0600,  
st_size=1944, ...}) = 0
lstat64("/backup/disk1/ns2_maildir/rdiff-backup-data/increments/ 
richard-group.com/david/cur/ 
1170338212.M12781P21813V0000000000000801I03130440_0.db1,S=3073:2,S",  
0xbf983bb0) = -1 ENOENT (No such file or directory)

Is it examining the directory to see what files exist in order to  
calculate the restore? Is it taking such a long time and running out  
of memory due to the number of files in the increments directory?  
There are 108440 files in this directory.

I suppose I could do this restore manually with gnu find, etc.  
However, if would be good if there's a way to get rdiff-backup to  
work here.

Any comments or suggestions?


--
Kurt Yoder
The Richard Group
v: 703.584.5803
f: 703.991.8990