[CRIU] can criu restore data after the reboot or using kexec ?

[Sanidhya Kashyap]

On Wed, Sep 2, 2015 at 8:32 AM, Pavel Emelyanov <xemul at parallels.com> wrote:
> On 09/02/2015 01:51 PM, lei yang wrote:
>> Hi experts,
>>
>> My target is using the nfs as root
>> root at ISG_WalnutCanyonII-2:~# cat /proc/cmdline
>> console=ttyS0,115200 root=/dev/nfs rw nfsroot=128.224.165.20:/export/pxeboot/vlm-boards/22574/rootfs rw ip=dhcp intel_iommu=off no_console_suspend=1 crashkernel=512M at 64M
>>
>> I follow below step:
>> ./test.sh  < /dev/null &> test.log &
>>  criu dump -v4 -o dump.log -D checkpoint/ -t 1264 --shell-job
>> then reboot
>> criu restore -d -D checkpoint/ -vvvv -o restore.log
>
> Try to use the --shell-job for restore too.
>

This case will work --shell-job for restore. But, on reboot, the
cgroups directory will be missing and you may need to think about how
to circumvent it. I have crude implementation which works (but pretty
old). I am working on porting to the latest criu tree.

>> it doesn't restore the test.sh, any help?
>>
>> if I use kexec, any point need to be careful ?
>
> For kexec case we have the WIP called RKU (rebootless kernel update) or KUP
> (kernel updater). Here's my talk at Plumbers last year [1], and And there
> also was an article from Sanidhya at [2], but it has moved :) so I'm putting
> the author in Cc.
>
> [1] http://www.slideshare.net/openvz/rku-dusseldorf
> [2] https://sslab.gtisc.gatech.edu/2015/rebootless-kernel-update.html
>

Well, the whole work is under submission. :)
I will release the whole code base after the notification (which is
16th Nov). Anyway, I will finish the kexec with memory binding
approach in a month. In case you need the code, I can share it with
you, but publicly it won't be possible for sometime.

Thanks,
Sanidhya

> -- Pavel