Re: [Qemu-devel] [PATCH 1/2] Add virtagent file system freeze/thaw

[Michael Roth <mdroth@linux.vnet.ibm.com>]

On 02/01/2011 08:41 AM, Stefan Hajnoczi wrote:
> On Tue, Feb 1, 2011 at 2:36 PM, Jes Sorensen<Jes.Sorensen@redhat.com>  wrote:
>> On 02/01/11 15:34, Stefan Hajnoczi wrote:
>>> On Tue, Feb 1, 2011 at 2:26 PM, Jes Sorensen<Jes.Sorensen@redhat.com>  wrote:
>>>> I have to admit you lost me here, where do you get that 500ms time from?
>>>> Is that the XMLRPC polling time or? I just used the example code from
>>>> other agent calls.
>>>
>>> 500 ms is made up.  I was thinking, "what would a reasonable polling
>>> interval be?" and picked a sub-second number.
>>>
>>> Can you explain how the timeout in fsfreeze can happen?  It's probably
>>> because I don't know the virtagent details.
>>
>> Ah ok.
>>
>>  From what I understand, the XMLRPC code is setup to timeout if the guest
>> doesn't reply within a certain amount of time. In that case, the caller
>> needs to poll to wait for the guest to complete the freeze. This really
>> should only happen if you have a guest with a large number of very large
>> file systems. I don't know how likely it is to happen in real life.
>
> Perhaps Michael can confirm that the freeze function continues to
> execute after timeout but the client is able to send fsstatus()
> requests?

Ahh, yeah there's the confusion: we only execute one RPC at a time, so a 
polling function for a previous RPC won't work unless that RPC is being 
done concurrently, via fork()ing or something and communicating status 
via some method of IPC.

I touched on possible approaches to dealing with this in the response I 
just sent to this patch.

>
> Stefan