[CRIU] Fw: Restore Error in Cgroup Systemd

Pavel Emelyanov xemul at parallels.com
Thu Jul 30 02:57:13 PDT 2015


On 07/30/2015 09:35 AM, Z F wrote:
> Hi,
> 
> I've tried on 1.6 and 2 problems encountered. 
> 1.  opts.manage_cgroups was set as CG_MODE_IGNORE.
> 2. After workaround above issue, migration still failed for same cause.
> Log attached.

Cyrill, it looks like we need a knob to ignore the systemd cgroups in criu code.
Can you prepare a patch for this?

> Best Regards
> Zhang Fan
> 
> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> *From:* Cyrill Gorcunov <gorcunov at gmail.com>
> *To:* Z F <full015zf at yahoo.com>
> *Cc:* "criu at openvz.org" <criu at openvz.org>
> *Sent:* Friday, July 24, 2015 4:41 PM
> *Subject:* Re: [CRIU] Fw: Restore Error in Cgroup Systemd
> 
> On Thu, Jul 23, 2015 at 06:46:34AM +0000, Z F wrote:
> 
> 
> 
>>   
>>    Hi,
>>   
>>    I am trying migrating a simple shell job between hosts and failed in
>>    restore as:
>>   
>>    (00.042759)  38289: Error (cgroup.c:908): cg: Can't move into
>>    systemd//user.slice/user-0.slice/session-741.scope/tasks (-1/-1): No such
>>    file or directory
>>    (00.043022) Error (cr-restore.c:1193): 38289 exited, status=1
>>    (00.044078) Error (cr-restore.c:1867): Restoring FAILED.
>>   
>>    The environments:
>>    [root at compute1 <mailto:root at compute1> criu-br-1.5-stable]# uname -a
>>    Linux compute1 3.12.44 #1 SMP Mon Jul 20 09:16:11 CST 2015 x86_64 x86_64
>>    x86_64 GNU/Linux
>>    [root at compute1 <mailto:root at compute1> criu-br-1.5-stable]# ./criu -V
>>    Version: 1.5.2
>>    [root at compute1 <mailto:root at compute1> criu-br-1.5-stable]# ./criu restore --images-dir
>>    /mnt/nfs/home/zf/criutest/dumps/ --shell-job -v4
>>    Restore log attached.
>>   
>>    The cause is that for each session in the system a directory created as
>>    /sys/fs/cgroup/system/user.slice/user-x.slice/session-y.scope/. While the
>>    source and destination have different sessions.
>>   
>>    Would you please help to solve the problem? Thank you very much.
> 
> 
> Please switch to criu 1.6 at least, there were a number of fixes for
> cgroups code. To migrate between systemd based nodes you need to prepare
> special slice for systemd on source node. See our vzctl tool which
> does such thing (https://src.openvz.org/projects/OVZ/repos/vzctl/browse)
> In short you need to prepare own slice for your container/application
> to be able to migrate it.
> 
> 
> 
> 
> 
> _______________________________________________
> CRIU mailing list
> CRIU at openvz.org
> https://lists.openvz.org/mailman/listinfo/criu
> 



More information about the CRIU mailing list