[one-users] datablocks continued

[Carsten.Friedrich at csiro.au]

How do you use the copy-on-write feature with Nebula? I use qcow images, but, judging by the time it takes, I'm sure currently the images are just normally copied to /.../var/X/images/ rather than using the copy-on-write feature.

Carsten

-----Original Message-----
From: users-bounces at lists.opennebula.org [mailto:users-bounces at lists.opennebula.org] On Behalf Of Javier Fontan
Sent: Friday, 3 December 2010 8:25
To: Steven Timm
Cc: users at lists.opennebula.org
Subject: Re: [one-users] datablocks continued

Hello,

Sparse files are supported by both ext* and nfs filesystems but
unfortunately scp command (used with tm_ssh) does not know what sparse
files are, this will make it copy the holes so the files in the end
will be bigger and it will transfer all the null data.

I do not recommend storing live images in nfs as it will become a
bottleneck to image copying (prolog and epilog times) and IO.

I see two ways to fix this. The easiest way is to modify the transfer
scripts to use a sparse file aware tool to copy images. One of those
is rsync and as it can use ssh communication method I think it is the
easiest to implement. You will have to change "scp" calls to rsync
calls with a special parameter so it deals with sparse files (--sparse
or -S). This can be done in one place, in tm_common.sh file there is a
list of commands that the scripts will use. The variable for SCP can
be changed to the path of rsync add also add the sparse flag:

SCP=/usr/bin/rsync --sparse

Hopefully this will solve the problems with sparse files and speed
image transfer.

The other way to do this is using qcow images. This image format is
not easily manageable with standard tools but provide some nice
features like compression, copy on write and sparse images without
needed filesystem support. Using this images will make scp behave
correctly as it is not an sparse file for the filesystem. Qcow is
really a nice format and it can be useful for more things other than
the sparse files problem you had but maybe it will take more time to
implement as you have to convert your images and modify your workflow.

More info on the format can be found here: http://linux.die.net/man/1/qemu-img

Bye

On Tue, Nov 30, 2010 at 5:07 PM, Steven Timm <timm at fnal.gov> wrote:
>
> Thanks to help from opennebula staff I was able to create
> a 250GB datablock in the image repository which initially
> is a sparse file.  i.e., ls returns a size of 250GB but
> du returns a size of only a few megabytes, the actual space
> used by the file system itself.  the datablock is persistent and is
> meant to be used only by one vm which will be operating all the time.
>
> However, I am now discovering that once I launch the vm which calls
> this datablock, the act of sending it through the tm_ssh adapters
> (i.e. transferring it via scp to the host node where the VM will run)
> makes the file not be sparse anymore, and it takes just
> as long to copy it there as if the file were not sparse.
> Also I am guessing that once I stop or shutdown this VM, the saving
> of the partition will take just as long.
> Is there any
> way to get around this?  can datablocks live on NFS? Can the
> whole image repository live on NFS?  If so, can the host nodes
> grab the images in question (OS or datablock) straight from the NFs server
> without sending them through the head node?
>
> Steve Timm
>
>
>
> ------------------------------------------------------------------
> Steven C. Timm, Ph.D  (630) 840-8525
> timm at fnal.gov  http://home.fnal.gov/~timm/
> Fermilab Computing Division, Scientific Computing Facilities,
> Grid Facilities Department, FermiGrid Services Group, Assistant Group
> Leader.
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>



-- 
Javier Fontan, Grid & Virtualization Technology Engineer/Researcher
DSA Research Group: http://dsa-research.org
Globus GridWay Metascheduler: http://www.GridWay.org
OpenNebula Virtual Infrastructure Engine: http://www.OpenNebula.org
_______________________________________________
Users mailing list
Users at lists.opennebula.org
http://lists.opennebula.org/listinfo.cgi/users-opennebula.org