[Linux-cluster] Snapshotting GFS and freezing

[Dirk H. Schulz]

Pasi Kärkkäinen schrieb:
> On Sat, Jan 23, 2010 at 11:12:54AM +0100, Dirk H. Schulz wrote:
>   
>> Ray Van Dolson schrieb:
>>     
>>> On Wed, Jan 20, 2010 at 06:14:02AM -0800, Dirk H. Schulz wrote:
>>>   
>>>       
>>>> That means calling
>>>>         sync
>>>>         echo 3 > /proc/sys/vm/drop_caches
>>>> inside the VM, right? Or is there anything more to do to flush everything?
>>>>
>>>> Dirk
>>>>     
>>>>         
>>> I would think the best way would be to actually pause the VM.  If you
>>> snapshot a VM while it's running, even with the above, how can you
>>> guarantee an application won't do something on the VM right as the
>>> snapshot occurs?
>>>   
>>>       
>> That is an important point, I am already pondering about it. Of course  
>> pausing the vm is the better option - if you can afford it.
>>
>> I am working systematically on an environment where the possibility of  
>> very high level availability vms is part of the design (e.g. 99,99 %).  
>> At the moment the only approach I can see is doing snapshots of a  
>> running vm and making those as reliable as possible.
>>     
>
> Pausing the VM doesn't really help. It doesn't guarangee in any way that 
> the applications (say MySQL) are in a safe/consistent state.
>
> You really need to coordinate the process with the applications, if you 
> realy need/want to make consistent backups with snapshots.
>   
Yes, of course, that is still another point in the list that has to be 
considered. But pausing your application (say, stop MySQL writing to the 
database) is not sufficient, you also have to write down buffers and 
caches of kernel, filesystem etc. to make the image as consistent as 
possible.

Dirk