[CRIU] [criu]launch page-server in swrk mode

[Dengguangxing]

在 2016/3/8 23:16, Nikita Spiridonov 写道:
> On Tue, 2016-03-08 at 20:16 +0800, Dengguangxing wrote:
>> Hi all,
>>
>> I am trying to lauch a page-server through criu swrk mode.
>>
>> The page-server will always exit with one pre-dump/dump call(with page-server opts).
>> I've tried to set keep_open flag as true in rpc request. and it just acts the same way.
>>
>> Am I wrong with the page-server swrk mode? How to lauch a page-server daemon in swrk mode
>> and make it survive serveral pre-dumps?
>>
>> Thanks!
>> Best Regards
>> Deng
>>
>> _______________________________________________
>> CRIU mailing list
>> CRIU at openvz.org
>> https://lists.openvz.org/mailman/listinfo/criu
> 
> Single swrk process must be able to handle multiple page-server
> requests, it works for me (at least with criu-1.8).
> 
> You can find example of usage in p.haul project. It is written on
> python, but use same rpc protocol.
> 
> criu exec in swrk mode (criu_conn.__init__ function):
> https://github.com/xemul/p.haul/blob/master/phaul/criu_api.py
> 
> page-server request creation (make_page_server_req function):
> https://github.com/xemul/p.haul/blob/master/phaul/criu_req.py
> 

Hi Nikita,

sorry for misleading. It seems phaul works this way(A, B are different hosts)
	step1 - A: page-server
	step2 - B: pre-dump --page-server
	step3 - A: page-server
	step4 - B: pre-dump --page-server
	   ... ...
	stepN - A: page-server
	stepN+1 - B: dump --page-server
I wonder if it is possible to supports:
	step1 - A: page-server
	step2 - B: pre-dump --page-server ...
	step3 - B: pre-dump --page-server ...
	   ... ...
	stepN - B: dump -page-server

As I know, the 2nd way is not supported, right?