Support C/R of frozen cgroup

[klesgidisold]

If I pause a container

docker pause <container_name>

and then try to checkpoint it

docker checkpoint <container_name>

checkpoint stalls...
From another terminal I am trying to unpause

docker unpause <container_name>

and it also stalls.

[xemul]

Seem to be Docker issue. @SaiedKazemi , @boucher would you please help here?

[SaiedKazemi]

Which Docker version are you using? Also, what is the container you're running? Can you send us the exact commands on how you started your server and the containers?

[klesgidisold]

Docker

Client:
 Version:      1.8.0-dev
 API version:  1.21
 Go version:   go1.4.2
 Git commit:   c74defa-dirty
 Built:        Wed Aug  5 23:15:11 UTC 2015
 OS/Arch:      linux/amd64
 Experimental: true

Server:
 Version:      1.8.0-dev
 API version:  1.21
 Go version:   go1.4.2
 Git commit:   c74defa-dirty
 Built:        Wed Aug  5 23:15:11 UTC 2015
 OS/Arch:      linux/amd64
 Experimental: true

I tested in a MySQL and a Postgre SQL (both official) containers. I run:

docker run -d --name <container_name> mysql:5.7

and then the commands that I specified on the first post.

I didn't do anything to start the docker server.
sudo service docker start (I'm not sure if you asking something else about it)

Thanks for the response..

[boucher]

Have you gotten checkpoint/restore to work at all on your system?

Out of curiosity, can I ask what the use case is for pausing the container before checkpointing it?

Also, could you try updating your build of Docker? There have been changes in my checkpoint/restore branch since the git hash shown in this version.

[klesgidisold]

Yes C/R is working on my system.

There isn't any use case. I was just messing around and noticed this attitude.. I decide to say it here if you haven't noticed it yet.

I am updating my current build of docker right now and I 'll let you know..

Thanks

[boucher]

Ok, thanks for letting us know. I don't actually know much about how docker
pause works, so it will require some looking into.

On Fri, Aug 14, 2015 at 9:20 AM, Kyriakos Lesgidis <[email protected]

wrote:

Yes C/R is working on my system.

There isn't any use case. I was just messing around and noticed this
attitude.. I decide to say it here if you haven't noticed it yet.

I am updating my current build of docker right now and I 'll let you know..

Thanks

—
Reply to this email directly or view it on GitHub
https://github.com/xemul/criu/issues/20#issuecomment-131167307.

[SaiedKazemi]

@xemul @klesgidis
I just tested pausing (freezing) a shell process and then checkpointing it (i.e., no Docker) and CRIU hung. Here is what I did:

[Terminal A]
/bin/sh -c 'i=0; while true; do echo $i; i=$(expr $i + 1); sleep 3; done'

[Terminal B]
mkdir /sys/fs/cgroup/freezer/0
ps -efl | grep -w sh
echo 7042 > /sys/fs/cgroup/freezer/0/tasks
echo FROZEN > /sys/fs/cgroup/freezer/0/freezer.state
criu dump -v4 -D /tmp -o dump.log -j -t 7042

[Terminal C]
ps -efl | grep criu
strace -p 6449
Process 6449 attached
wait4(6355,

Unless I am missing something, it seems to be a CRIU issue. Any idea?

[avagin]

CRIU can't attach by ptrace() to frozen processes.
Recently, I added the freeze-cgroup option, it should be used for paused containers.

[xemul]

@avagin CRIU can attach with ptrace, this is what we do in your patches :) CRIU cannot run parasite in them, but this can be addressed by extending the freeze-seize-unfreeze-run logic you've introduced.

[xemul]

8b04551