-------- Forwarded Message --------
From: Eric Praetzel <praetzel@engmail.uwaterloo.ca>
To: matt@mattmccutchen.net
Subject: RSync and large amounts of data
Date: Tue, 4 Dec 2007 09:00:39 -0500 (EST)
Good day Mr McCutchen,
I've posted to the rsync list but nobody replied and you seem knowledgeable.
Can you think of any reason why for about 220G of files (about 4.5M files)
that rsync would not delete files which were deleted on the master?
We use rsync -a --delete via ssh and rsync port and yet when we delete
a user account on our main Solaris server the mirrored arrays (Centos 5.0
and Solaris 8) don't delete the user accounts.
I've actually lost confidence and have resorted to testing to figure out
just how far I can trust rsync. It appears as if it's only for the root
directory that sub directories are not deleted if they're removed from
the master.
We're just about to switch over from our Sun servers to Linux and it would
take days to do a fresh rsync - but incrementals don't seem reliable - at
least for deleting files and we've currently scheduled the deletion of about
1200 accounts a few days before the server switchover. Every testing I've
done shows that it copies over new/modified files; but it fails to delete
directories and files which are no longer on the master.
Do you have any clue what could be up?
I've been testing our web server (only about 40G of data) compared to the
main file server and for 40G of data everything works the way it should;
but clearly not for about 220G.
On the main server rsync takes about 300M - (there are 4 CPUs and 4G) so
it's not starved for ram and no users are allowed on the system and it's
strictly a Samba and NFS server - lightly loaded.
Thankyou,
- Eric Praetzel, ECE Dept, Unviersity of Waterloo
