Re: [sheepdog] Qemu Live Migration and Cache

2012-06-08 Thread Bastian Scholz

Am 2012-06-07 12:07, schrieb Yibin Shen:
On Thu, Jun 7, 2012 at 5:38 PM, Liu Yuan namei.u...@gmail.com 
wrote:

Maybe you refer to object cache?


Yes, sorry, mixed that two...

Without object cache, the answer is NO, the data will be known to 
the

cluster as soon as the request is completely.

With object cache enabled, probably 'nope' again, but I am not 100%
sure, live migration mechanism from QEMU will issue a flush request 
to
sheep and block for it, then will migrate memory to other machine. 
For

the last try with live migration, it did the right thing for me, but
you'd better test it with your own. When I have time, I'll try it 
again

and give you a solid answer.so
qemu will call sync operation before live migration,(nearly) all 
dirty

objects will be
flushed to sheepdog cluster.


Thanks for your infos.

I am using libvirt with virt-manager on top to manage my virtual
machines. With sheepdog 0.2.4 live migration works without problems
for month.

If the current sheepdog version will work for me, I make some more
migration tests and give a feedback.

Cheers

Bastian

--
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog


[sheepdog] Qemu Live Migration and Cache

2012-06-07 Thread Bastian Scholz

I have a question about migration a Qemu Virtual Machines from
one node to another...

Is there something I have to pay attention when using the
farm cache mechanisms?

For example, is it possible, that I ran my virtual Machine V
on host A, Start a live migration to host B, but some content
in the cache from host A would not known to the rest of the
cluster in time?

Cheers

Bastian
--
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog


Re: [sheepdog] Qemu Live Migration and Cache

2012-06-07 Thread Liu Yuan
On 06/07/2012 05:19 PM, Bastian Scholz wrote:

 I have a question about migration a Qemu Virtual Machines from
 one node to another...
 
 Is there something I have to pay attention when using the
 farm cache mechanisms?

 


Farm doesn't cache any data for regular IO. Instead, Farm caches
snapshot objects for recovery, which is totally irreverent to data
consistency of running VM.

Maybe you refer to object cache?

 For example, is it possible, that I ran my virtual Machine V
 on host A, Start a live migration to host B, but some content
 in the cache from host A would not known to the rest of the
 cluster in time?
 


Without object cache, the answer is NO, the data will be known to the
cluster as soon as the request is completely.

With object cache enabled, probably 'nope' again, but I am not 100%
sure, live migration mechanism from QEMU will issue a flush request to
sheep and block for it, then will migrate memory to other machine. For
the last try with live migration, it did the right thing for me, but
you'd better test it with your own. When I have time, I'll try it again
and give you a solid answer.so

Thanks,
Yuan
-- 
sheepdog mailing list
sheepdog@lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog