Thank you both very much for your answer an explanation. after checking
with the ls -i command the two files are indeed on the same inode.
Le 01/07/2019 à 04:22, Adam Goryachev a écrit :
On 28/6/19 9:46 pm, Jean-Louis Biasini via BackupPC-users wrote:
Hi all,
I have a working installation on a centos7 server. I'm backing up 30+
linux server hosts with rsync. Since I was a bit surprised by the
growing space taken on the backup server, I started to investigate on
the pooling mechanism. So I started to create simple identical text
files on 2 servers to see if the file was showed as already in pool
while backing up the second server after having back up the first.
First I just created 2 times the same text file (ie same content,
same md5sum, same permission, same selinux but different creating
time) second I created it on the first server then rsync it to the
second to have it absolutely identical (rsync -aAXv). In both case
the file is showed as created by the second server's backup. Then I
tried a bigger file created with fallocate -l 2M test.img and rsynced
it the same way. In all case my file is created again on the second
backup. I also checked the hard link limit that I increased from
32000 to 64000 (ext4 file-system here) with no improvement. Am I
missing something?
You didn't mention which version of backuppc you are using, so I can't
be sure, but from memory, with BPC 3.x that is the correct/expected
behaviour. However, while the file is transferred from the second host
to the BPC server, it *will* get pooled correctly, so there will only
be a single on-disk version of the file. You can check this by
confirming the hard link entry on the BPC server for the 2 hosts in
question.
eg:
ls -i /var/lib/backuppc/pc/<host1>/123/f%2f/ftest.txt
/var/lib/backuppc/pc/<host2>/124/f%2f/ftest.txt
If the inode number is the same for both files, then they are
correctly de-duplicated/pooled.
This might be the same for BPC 4.x, although I recall (and have
confirmed ages ago) that BPC 4.x doesn't transfer the file if the
duplicate file is located on the same host. So, it may still transfer
the file in your example (different hosts), but again should not be
duplicated in the pool (I don't know how to verify this on BPC 4.x).
Maybe something like this:
/usr/lib/backuppc/bin/BackupPC_attribPrint
/var/lib/backuppc/pc/<host1>/123/f%2f/attrib_*
Check the inode entry, and do the same for your second host, make sure
the inode entry is the same, and I assume this would mean that the
pooling is working correctly.
PS, just because the xferlog shows a file is transferred, doesn't mean
it isn't pooled after BPC receives the full file, and confirms the
content is a match for the other file it already has.
Regards,
Adam
_______________________________________________
BackupPC-users mailing list
BackupPC-users@lists.sourceforge.net
List: https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki: http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/
