VERY frustrated with FreeBSD/UFS stability - please help or comment...

[Bill Vermillion]

The time has come the Dag-Erling Sm??rgrav said, to talk of many things
but all that was heard on Tue, May 22, 2007 at 13:39 was whether
pigs have wings - or:

> Gore Jarold <gore_jarold at yahoo.com> writes:
> > Specifically, I have private departmental fileservers that other
> > fileservers rsync to using Mike Rubel-style rsync snapshots:

> > http://www.mikerubel.org/computers/rsync_snapshots/

> > This means that the remote system runs a script like this:

> > ssh user at host rm -rf backup.2
> > ssh user at host mv backup.1 backup.2
> > ssh user at host cp -al backup.0 backup.1
> > rsync /files user at host:/backup.0

> This is extremely inefficient, as you have discovered.

> Speaking in the abstract, what you want to do every day is the
> following:

> client1% rsync --archive --delete /vol server:/backup/client1
> client2% rsync --archive --delete /vol server:/backup/client2
> server% for vol in /backup/* ; do mksnap_ffs $vol $vol/.snap/`date` ; done

> No copying or deleting; you take a snapshot when the rsync job
> is done, and the next day you rsync again to the same directory;
> only what has actually changed will be transferred, and there
> is no need to create and populate full copies of each directory
> tree every time.

And one other way to 'copy' files/directories >>IF<< they are
on the same file system, is to use cpio with the -pdlm option.

All that does is build another directory with all files in the
first liked statically to the second.  Then you just 'rm' the files
in the first.  Since there is NO COPYING - this is quick, won't 
scatter files around as they remain where they were originally but
with just a new directory pointing to them.

Probably not the best way for something done regularly but
a very quick and efficient way to move files around in a system
with very low overhead and is quite fast.  It also has the plus
for the new directory is that it gets you a clean directory, which
can be a perfomance gain if the original had lots of file, many of
which were deleted.

> The easiest way to do this (if you're not afraid to run
> experimental code) is to use ZFS on the server, as it lets you
> easily create separate file systems for each client, and creates
> and maintains snapshots far more cheaply than FFS.

I've been watching the ZFS and I'm going to wait awhile, but it
will be awhile before I put in on my servers.

Bill

-- 
Bill Vermillion - bv @ wjv . com