<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 18/02/16 18:17, Ján Tomko wrote:<br>
</div>
<blockquote cite="mid:20160218151753.GD22539@dnr.brq.redhat.com"
type="cite">
<pre wrap="">On Thu, Feb 18, 2016 at 05:04:16PM +0300, Maxim Nestratov wrote:
</pre>
<blockquote type="cite">
<pre wrap="">18.02.2016 16:46, Ján Tomko пишет:
</pre>
<blockquote type="cite">
<pre wrap="">On Wed, Feb 17, 2016 at 02:40:05PM +0300, Olga Krishtal wrote:
</pre>
<blockquote type="cite">
<pre wrap="">To update information about ploop volumes inside the a single pool we need
to be sure that it is the ploop directory and not some other directory.
Ploop volume directory obligatory contains root.hds - image file and disk
descriptor - DiskDescriptor.xml. If path to a volume is a path to some
directory, we check the existance of this files.
</pre>
</blockquote>
<pre wrap="">With each ploop volume being a directory with a ploop disk image and the
XML, I think they deserve a separate pool type.
</pre>
</blockquote>
</blockquote>
<pre wrap="">
On second thought, the pool is still just a directory, it's the volumes
that are different, so it does belong in the directory-based pools.</pre>
</blockquote>
<meta charset="utf-8">
<span style="color: rgb(37, 37, 37); font-family: sans-serif;
font-size: 14px; font-style: normal; font-variant: normal;
font-weight: normal; letter-spacing: normal; line-height: 22.4px;
orphans: auto; text-align: start; text-indent: 0px;
text-transform: none; white-space: normal; widows: 1;
word-spacing: 0px; -webkit-text-stroke-width: 0px; display: inline
!important; float: none; background-color: rgb(255, 255, 255);">The
main idea of ploop is to have an image file, use it as a block
device,<br>
and create and use a file system on that device. It is looks like
loop device.<br>
However, we do need </span>DiskDescriptor.xml store at the same
directory<br>
that root.hds (file, that contains ploop).<br>
<meta charset="utf-8">
<br>
<blockquote cite="mid:20160218151753.GD22539@dnr.brq.redhat.com"
type="cite">
<pre wrap="">
That leaves the mixing of the ploop volumes with the ploop disk images,
don't we need a new STORAGE_VOL type?</pre>
</blockquote>
At the beginning I also had thoughts about it, but:<br>
1) New STORAGE_VOL
type will be exclusively for us. No one else will use it.<br>
So, we have volume time only for one format.<br>
2) We can avoid such situation, because our ploop format looks a bit
like qcow format.<br>
(Of course they have have different internal structure, but still,
the only difference in such case -<br>
DiskDescriptor.xml, that can not be stored in the same file)<br>
3) And we are able to work with STORAGE_VOL_FILE, because we are a
file upon loop device.<br>
<br>
<blockquote cite="mid:20160218151753.GD22539@dnr.brq.redhat.com"
type="cite">
<pre wrap="">
</pre>
<blockquote type="cite">
<blockquote type="cite">
<pre wrap="">The ploop image (root.hds) could be detected as such by the fs pool,
but creating and deleting the directories feels out of place in this
backend.
</pre>
</blockquote>
<pre wrap="">
Actually one of the main intention of implementing ploop disk support in
storage pool was
ability to make it possible to work with existing storage pools like
NFS, DIR, FS.
</pre>
</blockquote>
<pre wrap="">
The 'image in a directory' still won't work with older libvirtd.
</pre>
<blockquote type="cite">
<pre wrap="">Creating a
separate storage pool makes it impossible. Also, our future plans to
expand storage pools
with new pool type like VZ storage and CEPH FS becomes worthless either.
</pre>
</blockquote>
<pre wrap="">
Jan
</pre>
</blockquote>
<br>
</body>
</html>