Re: [Qemu-devel] migration: broken ram_save_pending

[Alexey Kardashevskiy <aik@ozlabs.ru>]

On 02/05/2014 01:00 AM, Paolo Bonzini wrote:
> Il 04/02/2014 13:16, Alexey Kardashevskiy ha scritto:
>> On 02/04/2014 11:07 PM, Paolo Bonzini wrote:
>>> Il 04/02/2014 12:59, Alexey Kardashevskiy ha scritto:
>>>>>> With the default throttling of 32 MiB/s, bandwidth must be something
>>>>>> like
>>>>>> 33000 (expressed in bytes/ms) with the default settings, and then
>>>>> max_size
>>>>>> should be 33000*3*10^9 / 10^6 = 6000000.  Where is my computation wrong?
>>>>
>>>> migrate_max_downtime() = 30000000 = 3*10^7.
>>>
>>> Oops, that's the mistake.
>>
>> Make a patch? :)
> 
> I mean, my mistake. :)  I assumed 3000 ms = 3*10^9.
> 
> 30 ms is too little, but 3000 ms is probably too much for a default.

So - the default is bad and we need a patch (to make it 300ms), no?

> 
>>>> When the migration is in iterating stage, bandwidth is a speed in last
>>>> 100ms which is usually 5 blocks 250KB each so it is
>>>> 1250000/100=12500bytes/s and max_size=12500*30000000/10^6=375000 which is
>>>> less than the last chunk is.
>>>
>>> Perhaps our default maximum downtime is too low.  30 ms doesn't seem
>>> achievable in practice with 32 MiB/s bandwidth.  Just making it 300 ms or
>>> so should fix your problem.
>>
>> Well, it will fix it in my particular case but in a long run this does not
>> feel like a fix - there should be a way for migration_thread() to know that
>> ram_save_iterate() sent all dirty pages it had to send, no?
> 
> No, because new pages might be dirtied while ram_save_iterate() was running.


I do not get it, sorry. In my example the ram_save_iterate() sends
everything in one go but its caller thinks that it did not and tries again.
This is is not because something got dirty in between, this is only because
of sending zero pages as 8+1 bytes chunk (not as 4096+1 bytes).


-- 
Alexey