4:14 a.m.
This is a multi-part message in MIME format.
--------------090705080902000503070107
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
On 03/18/2013 04:43 PM, Ayal Baron wrote:
----- Original Message -----
> Mark Wu :
>> On Sun 17 Mar 2013 10:12:55 PM CST, Ayal Baron wrote:
>>>
>>> ----- Original Message -----
>>>> Hi guys,
>>>>
>>>> Currently, ISO domain is only supported on NFS storage. It could
>>>> improve the ease of use if it allows other types
>>>> of file based storage to store ISO images. After an
>>>> investigation, I
>>>> found there's not any restriction on this idea.
>>>> So the whole work is removing the limitation on engine side. That
>>>> means engine should allow ISO domain could
>>>> have different storage type from the data center it's attached,
>>>> like
>>>> what we do with nfs ISO domain in SAN DC.
>>>>
>>>> I start this idea with localfs. I know local storage can't be
>>>> seen in
>>>> cluster level. But it also provides a choice if no
>>>> NFS available. VMs can be created on the host which has the ISO
>>>> repo,
>>>> and then be migrated to any other host in the cluster.
>>>> I have done the initial patches: allow creation ISO domain on
>>>> localfs
>>>> [1] and support import ISO domain on localfs [2]
>>>> I don't have much experience in java/j2ee/web development and
>>>> engine
>>>> architecture. The patches just work for me.
>>>> I am not sure if it will bring some potential problems. So any
>>>> feedback on the patch or the idea will be appreciated very much.
>>> Haven't looked at the patches yet, but wrt the idea, I agree on
>>> the
>>> need (being able to attach ISOs from anywhere and not just nfs)
>>> but I
>>> think the way to do this should be by getting rid of the ISO
>>> domain
>>> type altogether.
>> I think ISO domain on localfs is useful for a simple setup or demo,
>> such as oVirt all-in-one.
>>
>>> Basically what we need is:
>>> 1. a way to connect to file based storage (let's leave block aside
>>> for now) - this already exists via the connectStorageServer verb
>>> 2. a way to list and present a file system tree in gui (give an
>>> arbitrary path to vdsm and list content) and possibly filter
>>> results
>>> by type (vfd, iso) - does not exist today. Possibly some security
>>> aspects here that need hashing out.
>>> 3. a way to specify a path to a file when attaching an iso/vfd to
>>> a
>>> VM - this is the way it works today
>>>
>>> This would devoid the need for isoUploader and allow users to
>>> simply
>>> manage an nfs export with files.
>>> Next step would be to make connectStorageServer support httpfs [1]
>>> and then we'd be able to mount ISOs directly over http (hopefully
>>> this would be sufficient to support ISOs stored on S3, swift,
>>> glance,
>>> etc).
>> Actually, we could use the qemu curl backend image support
>> directly.
>> That means we don't need mount the place storing ISO images. We can
>> just maintain a list of ISO image with its link, which could be
>> http,
>> ftp and ssh.
> That will be fine to start a VM on a existing extern ISO image. I
> also
> would like to maintain a ISO image cache on the local host to avoid
> to
> re-streaming the ISO image from the ISO image repositories every
> time.
> That will be helpful for people who is suffered from the network
> bottleneck.
We have a similar requirement for glance support (not limited to ISOs, rather to all
read-only images) so that should be tackled with a broader solution.
In that case,
I suppose qemu's copy-on-read + httpfs (you mentioned
above) could help. It can avoid accessing the same backing file sectors
repeatedly.
But the copied content can't be shared by multiple VM using the same
backend image.
>>> [1] http://httpfs.sourceforge.net/
>>>
>>>>
>>>> Mark.
>>>>
>>>> [1] http://gerrit.ovirt.org/#/c/12687/
>>>> [2] http://gerrit.ovirt.org/#/c/12916/
>>>>
>>>> _______________________________________________
>>>> Engine-devel mailing list
>>>> Engine-devel(a)ovirt.org
>>>> http://lists.ovirt.org/mailman/listinfo/engine-devel
>>>>
>>
>> _______________________________________________
>> Engine-devel mailing list
>> Engine-devel(a)ovirt.org
>> http://lists.ovirt.org/mailman/listinfo/engine-devel
>>
>
> --
> ---
> 舒明 Shu Ming
> Open Virtualization Engineerning; CSTL, IBM Corp.
> Tel: 86-10-82451626 Tieline: 9051626 E-mail: shuming(a)cn.ibm.com or
> shuming(a)linux.vnet.ibm.com
> Address: 3/F Ring Building, ZhongGuanCun Software Park, Haidian
> District, Beijing 100193, PRC
>
>
>
--------------090705080902000503070107
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit
<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 03/18/2013 04:43 PM, Ayal Baron
wrote:<br>
</div>
<blockquote
cite="mid:1599625604.9482831.1363596188030.JavaMail.root@redhat.com"
type="cite">
<pre wrap="">
----- Original Message -----
</pre>
<blockquote type="cite">
<pre wrap="">Mark Wu :
</pre>
<blockquote type="cite">
<pre wrap="">On Sun 17 Mar 2013 10:12:55 PM CST, Ayal Baron
wrote:
</pre>
<blockquote type="cite">
<pre wrap="">
----- Original Message -----
</pre>
<blockquote type="cite">
<pre wrap="">
Hi guys,
Currently, ISO domain is only supported on NFS storage. It could
improve the ease of use if it allows other types
of file based storage to store ISO images. After an
investigation, I
found there's not any restriction on this idea.
So the whole work is removing the limitation on engine side. That
means engine should allow ISO domain could
have different storage type from the data center it's attached,
like
what we do with nfs ISO domain in SAN DC.
I start this idea with localfs. I know local storage can't be
seen in
cluster level. But it also provides a choice if no
NFS available. VMs can be created on the host which has the ISO
repo,
and then be migrated to any other host in the cluster.
I have done the initial patches: allow creation ISO domain on
localfs
[1] and support import ISO domain on localfs [2]
I don't have much experience in java/j2ee/web development and
engine
architecture. The patches just work for me.
I am not sure if it will bring some potential problems. So any
feedback on the patch or the idea will be appreciated very much.
</pre>
</blockquote>
<pre wrap="">
Haven't looked at the patches yet, but wrt the idea, I agree on
the
need (being able to attach ISOs from anywhere and not just nfs)
but I
think the way to do this should be by getting rid of the ISO
domain
type altogether.
</pre>
</blockquote>
<pre wrap="">
I think ISO domain on localfs is useful for a simple setup or demo,
such as oVirt all-in-one.
</pre>
<blockquote type="cite">
<pre wrap="">Basically what we need is:
1. a way to connect to file based storage (let's leave block aside
for now) - this already exists via the connectStorageServer verb
2. a way to list and present a file system tree in gui (give an
arbitrary path to vdsm and list content) and possibly filter
results
by type (vfd, iso) - does not exist today. Possibly some security
aspects here that need hashing out.
3. a way to specify a path to a file when attaching an iso/vfd to
a
VM - this is the way it works today
This would devoid the need for isoUploader and allow users to
simply
manage an nfs export with files.
Next step would be to make connectStorageServer support httpfs [1]
and then we'd be able to mount ISOs directly over http (hopefully
this would be sufficient to support ISOs stored on S3, swift,
glance,
etc).
</pre>
</blockquote>
<pre wrap="">
Actually, we could use the qemu curl backend image support
directly.
That means we don't need mount the place storing ISO images. We can
just maintain a list of ISO image with its link, which could be
http,
ftp and ssh.
</pre>
</blockquote>
<pre wrap="">
That will be fine to start a VM on a existing extern ISO image. I
also
would like to maintain a ISO image cache on the local host to avoid
to
re-streaming the ISO image from the ISO image repositories every
time.
That will be helpful for people who is suffered from the network
bottleneck.
</pre>
</blockquote>
<pre wrap="">
We have a similar requirement for glance support (not limited to ISOs, rather to all
read-only images) so that should be tackled with a broader solution.</pre>
</blockquote>
In that case, I suppose qemu's copy-on-read + httpfs (you mentioned
above) could help. It
<meta http-equiv="content-type" content="text/html;
charset=UTF-8">
can avoid accessing the same backing file sectors repeatedly. <br>
But the copied content can't be shared by multiple VM using the same
backend image.<br>
<br>
<blockquote
cite="mid:1599625604.9482831.1363596188030.JavaMail.root@redhat.com"
type="cite">
<pre wrap="">
</pre>
<blockquote type="cite">
<pre wrap="">
</pre>
<blockquote type="cite">
<pre wrap="">
</pre>
<blockquote type="cite">
<pre wrap="">
[1] <a class="moz-txt-link-freetext"
href="http://httpfs.sourceforge.net/">http://httpfs.sourcefo...
</pre>
<blockquote type="cite">
<pre wrap="">
Mark.
[1] <a class="moz-txt-link-freetext"
href="http://gerrit.ovirt.org/#/c/12687/">http://gerrit.ovir...
[2] <a class="moz-txt-link-freetext"
href="http://gerrit.ovirt.org/#/c/12916/">http://gerrit.ovir...
_______________________________________________
Engine-devel mailing list
<a class="moz-txt-link-abbreviated"
href="mailto:Engine-devel@ovirt.org">Engine-devel@ovirt.org</a>
<a class="moz-txt-link-freetext"
href="http://lists.ovirt.org/mailman/listinfo/engine-devel">...
</pre>
</blockquote>
<pre wrap="">
</pre>
</blockquote>
<pre wrap="">
_______________________________________________
Engine-devel mailing list
<a class="moz-txt-link-abbreviated"
href="mailto:Engine-devel@ovirt.org">Engine-devel@ovirt.org</a>
<a class="moz-txt-link-freetext"
href="http://lists.ovirt.org/mailman/listinfo/engine-devel">...
</pre>
</blockquote>
<pre wrap="">
--
---
舒明 Shu Ming
Open Virtualization Engineerning; CSTL, IBM Corp.
Tel: 86-10-82451626 Tieline: 9051626 E-mail: <a
class="moz-txt-link-abbreviated"
href="mailto:shuming@cn.ibm.com">shuming@cn.ibm.com</a> or
<a class="moz-txt-link-abbreviated"
href="mailto:shuming@linux.vnet.ibm.com">shuming@linux.vnet.ibm.com</a>
Address: 3/F Ring Building, ZhongGuanCun Software Park, Haidian
District, Beijing 100193, PRC
</pre>
</blockquote>
<pre wrap="">
</pre>
</blockquote>
<br>
</body>
</html>
--------------090705080902000503070107--