On Thu, Mar 9, 2017 at 11:39 AM, Nathanaël Blanchet <blanchet@abes.fr> wrote:

Le 09/03/2017 à 10:25, Gianluca Cecchi a écrit :

On Wed, Mar 8, 2017 at 6:05 PM, Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:

NOTE: during the snapshot creation I see in web admin console the VM in paused state and also not responsive in both console and ssh session.

After a couple of seconds it comes back and as a confirmation I see this in its messages:

Mar 8 17:38:57 T-ORACLE73 chronyd[616]: System clock wrong by 19.077230 seconds, adjustment started

Is this expected?

Possibly the default changed at some point in time, so that now it saves memory and so this implies pause of VM

Saving memory is essential in some apàplications like DB, so you won't bypass vm pauses for such a stuff

Yes, indeed, the important thing is to have an option so that you can set it True or False, depending on the VM you are saving, the application that is running isnide it and the way you want to do backup of the application.

Nevertheless, RDBMS and also other applications often have some mechanism to be "frozen in a consistent state" so that you can save what you have on disk without need to save memory to have a consistent backup.

Oracle for example has functionality to be put in "backup mode" where you issue "begin backup" before the snapshot and "end backup" right after snapshot completion.

I see that POstgreSQL has similar functionality (not tested myself):

https://www.postgresql.org/docs/9.4/static/continuous-archiving.html#BACKUP-LOWLEVEL-BASE-BACKUP

and the same for other ones.

Gianluca