[Top] [All Lists]

Re: clone of filesystem across network preserving ionodes

To: "Meij, Henk" <hmeij@xxxxxxxxxxxx>
Subject: Re: clone of filesystem across network preserving ionodes
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Tue, 4 Nov 2014 08:56:04 +1100
Cc: Eric Sandeen <sandeen@xxxxxxxxxxx>, "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <8688BD11DAC0574AA90295127E9E9F4AC047F22D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <8688BD11DAC0574AA90295127E9E9F4AC047F1CA@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <5457E813.6030606@xxxxxxxxxxx> <8688BD11DAC0574AA90295127E9E9F4AC047F22D@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Mon, Nov 03, 2014 at 09:10:10PM +0000, Meij, Henk wrote:
> inodes, yes, yes.  I tried xfs_copy but ran into a problem.
> server1 and server2 each have 4x28T partitions and on server1
> writing the file to one empty partition mounted never finishes,
> hangs at 90%...presumably because it runs out of space (file size
> == sdbx).

Sure, but it's a sparse copy. It only copies the allocated blocks in
the filesystem, so the actual space required is the used space int
eh filesystem (i.e. what df reports as used).

> there is no "skip empty inodes" option and perhaps there
> can't be...

Of course not - you're wanting identical inode numbers on
either end, so your only option is and identical copy. Otherwise
you'd use xfsdump/xfsrestore to skip empty inodes...

> and I have nothing larger. I guess I could attempt
> making sdb4 slightly larger by reducing sdb1-3.

Exactly why do you need an *identical* copy of 28TB filesystems?
What is the problem with inode numbers being different?  And that
begs the question: if you need the filesystems to be completely
identical yet exist on separate systems, then why aren't you using a
block layer construct designed for such operation (e.g. drdb)?


Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>