On 07/22/2015 01:21 AM, Pavel Gashev wrote:
On 22/07/15 00:11, "users-boun...@openvz.org on behalf of Kir Kolyshkin"
<users-boun...@openvz.org on behalf of k...@openvz.org> wrote:
Other "why not simfs" considerations are listed at
http://openvz.org/Ploop/Why#Before_ploop
That¹s good, but there is an issue with using ploops. If you want to
life-migrate a huge container to another node, it might take hours (or
even days). Life migration might fail at CPT stage. It¹s not an issue with
simfs since you can use the --keep-dst option, and continue migration in
offline mode. Using ploops you have to start migration from the beginning.
No, in fact you don't.
I am not quite sure about Virtuozzo tools, but with my tools (i.e. vzmigrate
from what is now called "legacy OpenVZ") you can use the same "-r no
--keep-dst"
flags as for simfs, with the very similar effect -- only top delta will
need to be re-copied
again, for all other deltas rsync will be used just in case of simfs.
Oh, and if the top delta is large, just create a snapshot before trying
live migration,
or use --snapshot switch to vzmigrate (currently, though, you'll need to
delete
the snapshot manually after migration -- this is something in my TODO list
I can't find time to implement).
In other words, a ³showstopper² with ploops is that you can¹t continue
stopped/failed migrations.
Thanks for detailed explanation, it makes sense, but definitely not for
OpenVZ -- I am pretty sure it works the way I described above.
_______________________________________________
Users mailing list
Users@openvz.org
https://lists.openvz.org/mailman/listinfo/users
