[one-users] Ceph and thin provision

[Campbell, Bill]

Yes, Dumpling supports format 2 images (I think Bobtail 0.56 was the first release that did). 

I'll be submitting my modified driver to the development team for inclusion/modification (ideally we should be able to select which format we want to use, so further modifications would be necessary) and hopefully it would be included in the next version. 

In the interim, I can share with you the drivers we are using, but be advised this would be UNSUPPORTED by the OpenNebula development/support team. It has been working rather well for us though. 

----- Original Message -----

From: "Kenneth" <kenneth at apolloglobal.net> 
To: "Bill Campbell" <bcampbell at axcess-financial.com> 
Cc: users at lists.opennebula.org 
Sent: Thursday, December 12, 2013 9:29:49 AM 
Subject: Re: [one-users] Ceph and thin provision 



This all is good news. And I think this will solve my problem of a bit slow (a few minutes) of deploying a VM, that is cloning is really time consuming. 

Although I really like this RBD format 2, I'm not quite adept yet on how to implement it in nebula. And my ceph version is dumpling 0.67, does it support rbd format 2? 

If you have any docs, I'd greatly appreciate it. Or rather I'm willing to wait a little longer, maybe on the next release of nebula(?), to make rbd format 2 to be the default format? 
--- 
Thanks,
Kenneth 
Apollo Global Corp. 


On 12/12/2013 09:48 PM, Campbell, Bill wrote: 


Ceph's RBD Format 2 images support the copy-on-write clones/snapshots for quick provisioning, where essentially the following happens: 
Snapshot of Image created --> Snapshot protected from deletion --> Clone image created from snapshot 
The protected snapshot acts as a base image for the clone, where only the additional data is stored in the clone. See more here: http://ceph.com/docs/master/rbd/rbd-snapshot/#layering 
For our environment here I have modified the included datastore/tm drivers for Ceph to take advantage of these format 2 images/layering for Non-Persistent images. It works rather well, and all image functions work appropriately for non-persistent images (save as, etc.). One note/requirement is to be using a newer Ceph release (recommend Dumpling or newer) and newer versions of QEMU/Libvirt (there were some bugs in older releases, but the versions from Ubuntu Cloud Archive for 12.04 work fine). I did submit them for improvement prior to the 4.0 release, but the simple format 1 images are the default currently for OpenNebula. 
I think this would be a good question for the developers. Would creating the option for Format 2 images (either in the image template as a parameter or on the Datastore as a configuration attribute) and then developing the DS/TM drivers further to accommodate this option be worth the effort? I can see use cases for both (separate images vs. cloned images having to rely on the base image), but cloned images are WAY faster to deploy. 
I have the basic code for format 2 images, I think the logic for looking up the parameter/attribute and then applying appropriate action should be rather simple. Could collaborate/share if you'd like. 
----- Original Message -----

From: "Kenneth" <kenneth at apolloglobal.net> 
To: users at lists.opennebula.org 
Sent: Thursday, December 12, 2013 6:11:15 AM 
Subject: Re: [one-users] Ceph and thin provision 


Yes, that is possible. But as I said, all my images were all preallocated as I haven't created any image from sunstone. 


--- 
Thanks,
Kenneth 
Apollo Global Corp. 


On 12/12/2013 06:25 PM, Michael wrote: 
<blockquote>

This doesn't appear to be the case, I've 2TB of images on Ceph and 380GB 
data reported by Ceph (760G after replication). All of these Ceph images 
were created through the Opennebula Sunstone template GUI.

-Michael

On 12/12/2013 09:11, Kenneth wrote: 

<blockquote>
I haven't tried creating a thin or thick provision in ceph rbd from scratch. So basically, I can say that a 100GB disk will consume 100GB RBD in ceph (of course it will be 200GB in ceph storage since ceph duplicates the disks by default). 


_______________________________________________
Users mailing list Users at lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org 

</blockquote>

_______________________________________________ 
Users mailing list 
Users at lists.opennebula.org 
http://lists.opennebula.org/listinfo.cgi/users-opennebula.org 

NOTICE: Protect the information in this message in accordance with the company's security policies. If you received this message in error, immediately notify the sender and destroy all copies. 
</blockquote>



NOTICE: Protect the information in this message in accordance with the company's security policies. If you received this message in error, immediately notify the sender and destroy all copies.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20131212/41b5bb0c/attachment-0002.htm>