[CRIU] Docker checkpoint: real-world workloads

[Alice Frosi]

Hi,

We have started to test docker checkpoint with a real-world workloads.

Until now we tried with acmeair-nodejs [1] with the microservice usecase.

This application creates 10 containers and they interact between each other with volumes and network

connection. The checkpoint war not successful. Here, the error message in restore:

(00.195135)      1: No iptables-9.img image

(00.195136)      1: Error (criu/include/image.h:144): BUG at criu/include/image.h:144

(00.195171) Error (criu/cr-restore.c:1284): 15638 killed by signal 127: Unknown signal 127

(00.315156) Error (criu/mount.c:2980): mnt: Can't remove the directory /tmp/.criu.mntns.C12otH: No such file or directory

(00.315160) Error (criu/cr-restore.c:2152): Restoring FAILED.

We tested with criu version:

Version: 3.4

GitID: v3.4-224-g3706af7b

1) Any ideas for what could have caused the error?

Additionally, we have two more general questions.

2) How should we perform checkpoint for applications with multiple container interaction?

3) For which use cases have you already tried the C/R?

We saw this overview in your wiki [2], but it seems to be not recently updated.

Regards

Alice Frosi

[1] https://github.com/blueperf/acmeair-nodejs/blob/master/README_Docker.md
[2] https://criu.org/What_software_is_supported