xfs
[Top] [All Lists]

Re: re[2]: enterprise-level experience

To: linux-xfs@xxxxxxxxxxx
Subject: Re: re[2]: enterprise-level experience
From: Marek.Les@xxxxxxxxxxxxxxx
Date: Fri, 7 Jun 2002 10:42:17 +0200
Sender: owner-linux-xfs@xxxxxxxxxxx
Hello,

>I agree with Simon.
>
>When you are talking about a couple of Terabytes, most Raid Arrays
effectively
>have their own LVM built in.

First of all, thanks to all for your responses.

Now, the storage solution I am talking about (unfortunately I'm not allowed
to be specific here, for the moment) is, what you called a high-end one,
and has some kind of LVM of it's own but from what I understood, it cannot
simply grow an existing logical volume. It has an internal RAID 5 and I've
been told that "it could have been done but it's not for the reason of
security". I can see some point in this, they want to maintain data
integrity (the way the device works internally is quite complex). So, what
happens is that when I want to "enlarge" the volume, I get (physically, in
the OS) a new device, like /dev/sdc. And now I'm on my own to grow a
filesystem, so I need to use software LVM..

>Generally, software LVM is just needed for 2 features:
>
>Snapshots:  This is available in many Raid systems, but it can be a fairly

>expensive option, so many people prefer software snapshots.  Since XFS
does not
>to my knowledge directly support snapshots, one is left to consider a
software
>LVM vs. paying for the hardware option.

As you say, this is a fairly expensive option :) But unfortunately, LVM1 is
deadlocking with xfs_freeze - anyone has any EVMS experience?

>Mirroring between Raid Arrays: Some truly paranoid people use LVM to
mirror 2
>standalone Raid Arrays to together.  This is pretty rare because it is
very
>expensive to do.  I doubt you are planning on doing this.

Yes, I won't do that.

Marek Les
Seznam.cz



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