[CRIU] Page Cache

[Diyu Zhou]

Hi Andrei,

Thanks for your reply. The use case I'm considering is to use CRIU for fault
tolerance. Suppose CRIU checkpoint the state of the process of interest and
send the state to a backup machine. The state CRIU checkpoint must include
the
page cache, otherwise, during fail over, the restored state is not
consistent
with the disk state.

I agree with you that the probably best way to handle page cache is to
perform
sync before checkpoint. I just want to confirm that CRIU does not handle
page
cache and thanks for your help!

Best,
Diyu

On Thu, Jun 13, 2019 at 11:07 AM Andrei Vagin <avagin at gmail.com> wrote:

> Hi Diyu,
>
> CRIU handles runtime states of processes and it does nothing with
> persistent file systems. If you dump and restore processes on the same
> host without moving files, the Linux kernel will write all dirty pages
> to disks. If you migrate processes to another host, you need to sync
> all dirty pages to disks before migrating a persistent storage to the
> remote host. In many cases, it will not requires any additional
> actions from your end, it will be done implicitly.
>
> Maybe you can describe your use cases with more details.  Right now, I
> don't understand why criu should care about this.
>
> Thanks,
> Andrei
>
> On Wed, Jun 12, 2019 at 4:30 PM Diyu Zhou <zhoudiyupku at gmail.com> wrote:
> >
> > Hey CRIU,
> >
> > Does CRIU handle dirty pages in page cache during checkpointing? It
> seems to me
> > the code is not there in the main stream commit. Are there code for this
> in
> > other branches by any chance?
> >
> > Thanks in advance for your help!
> >
> > Thanks,
> > Diyu
> > _______________________________________________
> > CRIU mailing list
> > CRIU at openvz.org
> > https://lists.openvz.org/mailman/listinfo/criu
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/criu/attachments/20190614/84909177/attachment-0001.html>