VVR snap and db backup mode

 Hi Marco,

Yes, you always need to put a database into a backup-mode (or some kind of consistent mode) before taking a snapshot.

As you know, this is complicated in VVR because the secondary may be a few minutes (or whatever) behind if you have asynchronous replication.

That said, VVR has a feature, called IBC (In-Band Control), which allows you to put the database into backup mode on the Primary site, then send an IBC command to do a "snapshot", and then take it out of backup mode.  When the secondary "catches up" to when the database was in a consistent state, the IBC does the snapshot in the secondary side. This brilliant feature allows you to not have to wait or pause replication or any of that other stuff.

See the documentation on IBC feature (or attend the VVR course, as we talk about exactly that functionality and have you use it in the labs).

Hope that helps.

Cheers,
Tomer

 

Thanks Tomer,

but I wannt to understand deeply:

• I did snapshot on secondary only pausing the secondary rlink and the primary rlink the data was consistent and behind of 16GB (the data repplicated is 7TB) and we started the db from snapshot volumes on secondary without problems, so is suggested to put the db in backupmode or is mandatory? was I luky? :-)
• in a scenario without snapshot, only VVR for volumes of a db, with data consistent but not uptodate, if the primary site get destroyed, the secondary have to be able to start the db, do you agree?

The reason of my questions is that I want to be less intrusive on production service as possible


Thanks
Cheers
Marco

Here is a similar question:

Forget about VVR. If you have a single machine production database server that is running, without backup mode, would you use a snapshot to backup the data, of the primary database data?  Normally you would not make a snapshot or backup of a LIVE database becaues of concerns with consistency... specifically, the snapshot of the actual database data volume is a concern, since the database (without logs from the same point in time) would possibly not be recoverable.

In your example of the Secondary above, you are correct... the secondary can be recovered at any time when the Primary crashes, even though the database has been live... however that is because the Secondary has the database, logs etc all in a consistent point in time.

So, if your question is: can I pause the RLINK and then backup the Secondary, and use that as a valid backup. Yes.... however this backup: 1. after restored will have to perform a recovery as if it was data from a system crash, and 2. the Secondary has a small window of risk where you will fall more behind while you have paused replication (however since you say you do a snapshot then that window is probably pretty small). Just ensure that backing up the Secondary is done by backing up ALL data volumes including logs.

Hey, if your procedure works, it works. I can see how it does. However there is best practice which allows you to do this kind of thing, with more consistent backups, with minimal (no) risk to being able to takeover on the secondary site... however true the setup may be only slightly more complex (using IBC and snapshots; would require further testing by you). 

Thanks for your explanation.


Regards
Marco

Hi,
just to share our test results pausing the replication on secondary and taking a snapshot (without backup mode on db) the istance comed up (22TB)like after system crash.
Thanks to all
Regards
Marco