Error (criu/pstree.c:411): Current gid 1 intersects with pid (255) in images

[liuaifu]

I'm using criu 3.19 in the Docker container, the system in the container is Centos 7.9, and the host system is also Centos 7.9.
Use criu-ns dump -D ./ checkpoint -t xxx --leave-running -j dump created successfully , and using criu-ns restore -D ./ checkpoint -j restore process failed:
Error (criu/pstree.c:411): Current gid 1 intersects with pid (255) in images

Docker version is 20, start the container with the '--privileged' parameters, please help to see how to solve? Thanks.

[adrianreber]

This doesn't solve your problem, but CentOS 7 is end of life for over 1.5 years now. CentOS 7 never had all the necessary kernel patches to correctly run CRIU and CRIU 3.19 is also not the newest version.

Please update and try again. Also try to describe what you are trying to solve. Maybe we have same suggestions how to solve it differently.

[liuaifu]

@adrianreber Thank you for response.
The usage scenario looks something like this:

1. Program A starts B with boost::process::child, redirecting B's stdout and stderr;
2. Use criu-ns dump B;
3. Use criu-ns restore B from checkpoint.
   This works without problems on host os. But inside the container, dump appears normal, restore reports the error. If A starts B, detach B can also be dumped and restored normally in the container.

My requirement is that B can dump and restore normally in the container without detaching B from A. Is there a solution without upgrading the operating system and criu version?

[adrianreber]

Why is everything wrapped in criu-ns?

[rst0git]

@liuaifu Note that criu-ns is used mainly to address the PID mismatch problem by restoring into an empty PID namespace. If you are restoring in a container without any other process that may cause PID mismatch, you can use criu directly.

[liuaifu]

@adrianreber @rst0git That's because I want to implement multiple processes, and the original process needs to stay running.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.