[CRIU] p.haul: migrating process with TCP connection established

[Pavel Emelyanov]

On 04/09/2014 08:17 PM, Oscar Fernández wrote:
> Hello, responses inline.
> 
> 2014-04-09 17:44 GMT+02:00 Pavel Emelyanov <xemul at parallels.com <mailto:xemul at parallels.com>>:
> 
>     On 04/09/2014 07:26 PM, Oscar Fernández wrote:
>     > Hi all,
>     > I'm trying to migrate a process which has a TCP connection established. This process is the server side of a simple server-client TCP program.
>     >
>     > First of all I tried to dump and restore it using CRIU and it worked fine thanks to --tcp-established option.
>     >
>     > Now I want to move that process to another machine, so I'm using p.haul this time, but it fails on process restoration.
> 
>     Can you show the command you're doing it with?
> 
>  
> I'm using this command: p.haul/p.haul pid 3916 192.168.56.201 -v4

OK, then this is pid driver. As Andrey pointed out live-migrating of a single process
is tricky, since it's not clear what to do with the IP address. If you can take an
IP address with the process, then live-migraiton is possible, but this should be
properly coded in pid driver.

> 
>     > The content of criu_restore.log is:
>     >
>     > 3741: Error (sk-inet.c:514): Connected TCP socket in image
>     > Error (cr-restore.c:1036): 3741 exited, status=255
>     > Error (cr-restore.c:1579): Restoring FAILED.
> 
>     There's a -v<N> option to the p.haul that increases verbosity,
>     it will give us more info :)
> 
> 
> Please, find attached the criu_restore.log file.

Hm... Why did dump succeeded? It should have noticed that there was TCP socket alive
and abort the dump. Can you show the dump.log?

Thanks,
Pavel