[CRIU] Problem in Seizing Open File Descriptors?

[Pavel Emelyanov]

On 07/15/2014 08:30 PM, Saied Kazemi wrote:
> On Tue, Jul 15, 2014 at 8:08 AM, Pavel Emelyanov <xemul at parallels.com <mailto:xemul at parallels.com>> wrote:
> 
>     On 07/15/2014 06:49 PM, Pavel Emelyanov wrote:
>     > On 07/15/2014 06:34 PM, Saied Kazemi wrote:
>     >> Thanks for the quick feedback.  I am using CRIU 1.3-rc2 (at commit e1b56c8fa) with Docker version
>     >> 1.1.0 on Ubuntu 14.04 which does not provide mnt_id in /proc/pid/fdinfo/fd files.
>     >
>     > That can be a problem.
>     >
>     >> I will look into Docker source today.  Assuming that it does open /dev/null before moving into the
>     >> namespaces, can CRIU handle it?
>     >
>     > It should, but without this field this would fail just the
>     > way you observe with Docker :)
> 
>     After refreshing my memory and talking to Andrey (who wrote this code) the
>     answer is -- no. When we have multiple mount namespaces files opened from
>     outside are not dumped.
> 
> 
> Is this because this case cannot be technically supported, or is it because it has been
> a marginal case so far?

Because it has been a marginal case.

>  
> 
>     For /dev/null you could consider using the --evasive-devices option. It will
>     ask criu to dump the device's path in whichever location it's found. Not
>     sure how would this affect Docker though.
> 
> 
> The good news is that using --evasive-devices, it's possible to dump and restore the process.
> But, as you said, I am not sure yet how this would affect Docker.  Will look into it.

Great! :)
 
> 
>     > If you could confirm that on
>     > the kernel with mnt_id in fdinfo it works, that would be
>     > great. Meanwhile, I will try to make a test for this.
> 
> 
> Will have to find a kernel with mnt_id support and repeat the tests.  This may take a while.

Sure. The patch in question is the https://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=49d063cb3

Thanks,
Pavel