[CRIU] CRIU restored Docker IP address

[Rohit Upadhyaya]

Thanks Ross, with the branch you mentioned I was able to checkpoint,
restore and even connect to a container. But the restore and connect works
only if the IP assigned by Docker is the same on both the checkpoint host
and restore host. This usually means that I have to start the containers in
order so that the Docker daemon assigns the same IP in both places. I tried
to exec (privileged) into the machine and run 'ip addr dev eth0' but this
just gives me a 'RTNETLINK: Operation not permitted' error. Is there
anything else I could do?

Thanks,
Rohit

On Fri, Oct 23, 2015 at 12:40 PM, Rohit Upadhyaya <rohitju at cmu.edu> wrote:

> Ah I was on the wrong branch, my apologies. I am now able to see the
> checkpoint/restore commands and will test out connecting to a restored
> container which hopefully works. Thanks for the help!
>
> On Fri, Oct 23, 2015 at 12:11 PM, Ross Boucher <rboucher at gmail.com> wrote:
>
>> Are you sure you are on the right branch? I don't think that sounds
>> right. Sorry I'm on my phone I can look more closely in a bit.
>> On Fri, Oct 23, 2015 at 12:06 PM Rohit Upadhyaya <rohitju at cmu.edu> wrote:
>>
>>> Yup I see the "# WARNING! DOCKER_EXPERIMENTAL is set: building
>>> experimental features" message at the end. I'm using the binary
>>> bundles/1.7.0-dev/binary/docker-1.7.0-dev since the VERSION file in the
>>> huikang branch is 1.7.
>>>
>>> On Fri, Oct 23, 2015 at 10:46 AM, Ross Boucher <rboucher at gmail.com>
>>> wrote:
>>>
>>>> In your build log, towards the end, you should see this:
>>>>
>>>> # WARNING! DOCKER_EXPERIMENTAL is set: building experimental features
>>>>
>>>> Are you sure that you're then using the binary in
>>>> bundles/1.9.0-dev/binary/docker-1.9.0-dev?
>>>>
>>>> On Fri, Oct 23, 2015 at 10:45 AM Ross Boucher <rboucher at gmail.com>
>>>> wrote:
>>>>
>>>>> I build by just running this command:
>>>>>
>>>>> DOCKER_EXPERIMENTAL=1 make binary
>>>>>
>>>>>
>>>>> On Fri, Oct 23, 2015 at 10:43 AM Rohit Upadhyaya <rohitju at cmu.edu>
>>>>> wrote:
>>>>>
>>>>>> Thanks, Ross. Would you happen to know how I would build docker with
>>>>>> experimental features? I tried setting the DOCKER_EXPERIMENTAL=1
>>>>>> environment variable but the generated binary does not seem to have the
>>>>>> checkpoint and restore commands.
>>>>>> So, networking is most likely broken in that build. If you try
>>>>>> building from this branch, you might have better luck:
>>>>>>
>>>>>> https://github.com/huikang/docker/tree/fix-restore-network-cr-combined-1006
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Thu, Oct 22, 2015 at 6:50 PM Rohit Upadhyaya <rohitju at cmu.edu>
>>>>>> wrote:
>>>>>>
>>>>>>> Hi Ross,
>>>>>>>
>>>>>>> I am using Docker version 1.9.0 from the vagrant box here:
>>>>>>> https://vagrantcloud.com/kimh/boxes/criu and I followed the
>>>>>>> instructions here:
>>>>>>> http://blog.circleci.com/checkpoint-and-restore-docker-container-with-criu/.
>>>>>>> I am doing nothing more than running `docker checkpoint <id>` (with a TCP
>>>>>>> server running) followed by a `docker restore <id>`.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Rohit
>>>>>>>
>>>>>>> On Thu, Oct 22, 2015 at 6:45 PM, Ross Boucher <rboucher at gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi Rohit,
>>>>>>>>
>>>>>>>> Can you clarify a couple of things? Where did you get your binary
>>>>>>>> for checkpoint/restore in docker, what version is it? And, are you just
>>>>>>>> doing: docker checkpoint <id>; docker restore <id>; or something more
>>>>>>>> complicated?
>>>>>>>>
>>>>>>>> -Ross
>>>>>>>>
>>>>>>>> On Thu, Oct 22, 2015 at 6:42 PM Rohit Upadhyaya <rohitju at cmu.edu>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> I searched a bit on the CRIU archives and could find only one
>>>>>>>>> thread with a similar problem:
>>>>>>>>> https://lists.openvz.org/pipermail/criu/2015-April/020192.html
>>>>>>>>> .How would one use netns to connect a client to a restored container? I'm
>>>>>>>>> basically trying to reconnect to a server running on a restored Docker
>>>>>>>>> container and am kind of stuck on the IP issue.
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Rohit
>>>>>>>>>
>>>>>>>>> On Thu, Oct 22, 2015 at 10:04 AM, Rohit Upadhyaya <rohitju at cmu.edu
>>>>>>>>> > wrote:
>>>>>>>>>
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> I have been playing around with CRIU with Docker and am facing a
>>>>>>>>>> problem with the way a new IP address is assigned to a resumed container. I
>>>>>>>>>> have a TCP server running on the container and on restore a new IP address
>>>>>>>>>> is assigned and my client is unable to talk to the server (even on the new
>>>>>>>>>> IP). Looking at the restore log I find that the restore process restores
>>>>>>>>>> the socket descriptors with the old IP address. Is there any way to force
>>>>>>>>>> the restored container to run on the old IP?
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> CRIU mailing list
>>>>>>>>> CRIU at openvz.org
>>>>>>>>> https://lists.openvz.org/mailman/listinfo/criu
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/criu/attachments/20151110/0f7a5707/attachment.html>