[clug] a few LVM questions

[Alastair D'Silva]

We were running snapshots for VMs at work for a while, we ended up going
back to raw volumes for 2 reasons:

1. You are forced to guestimate the size of the snapshot required, if you
guess wrong, the snapshot will be offlined (with no chance of recovery,
although this was a couple of years ago)
2. There are only a restricted number of snapshots you can create, when you
hit this limit, you can't create any more, and your (lvm) scripts start
failing with bizarre error messages.

We were centralising our VM images on a fileserver, and exporting them via
iSCSI to diskless VMWare hosts.

In our next iteration, we are likely going to move to Solaris/ZFS, which
should alleviate these issues.

One other thing we did learn from this setup is that IOPs count when you are
serving VMs. We were running a RAID6 (Areca HW RAID) with 10 750GB drives,
and it was struggling to keep up. With our next generation setup, we plan to
use SSDs (probably more than one) as ZFS cache devices to allow the working
set of data to be served from them (with low latency), while the less used
data comes from the RAID proper. There is a good article explaining how it
works (and some anecdotal benchmarks) at
http://blogs.sun.com/brendan/entry/test


Cheers,

--
Alastair D'Silva           mob: 0423 762 819
Networking Consultant
New Millennium Networking  web: http://www.newmillennium.net.au
skype: alastair_dsilva     msn: alastair at d-silva.org
blog: http://alastair.d-silva.org


> -----Original Message-----
> From: linux-bounces at lists.samba.org [mailto:linux-
> bounces at lists.samba.org] On Behalf Of Andrew
> Sent: Sunday, 16 August 2009 6:37 PM
> To: linux at lists.samba.org
> Subject: [clug] a few LVM questions
> 
> Hi All,
> 
> I'm running a few virtual machines on a LVM volume (and I have some
> space on the base partition for snapshots).
> 
> I have been allocating 3GB  for a snapshot (and then rsync the vm's to
> another server).  I was wondering if there is any way to work out how
> much of this space is actually being used? (I googled and found a few
> other people with the same question.... but frustratingly all the
> howtos
> I found didn't answer this question) Likewise I am curious to know what
> happens if it fills up (what errors should I check for, and where)?
> 
> I also noticed that when I create the snapshot it blocks all writing
> for
> the virtual machines when the snapshot is being created (which does
> create some problems from time to time), googling around said that this
> wasn't a problem in earlier kernels, but some other users have noticed
> this as well.   Anyhow, was wondering if anyone else had experienced
> this?
> 
> 
> Other info:  Running Debian Lenny (64 bit, 2.6.30 kernel) + 4 x 15,000
> 146GB SAS Drives in RAID 5.
> 
> 
> Cheers,
> Andrew
> --
> linux mailing list
> linux at lists.samba.org
> https://lists.samba.org/mailman/listinfo/linux