Discussions specific to the VMware vSphere hypervisor
Post Reply
adb98
Enthusiast
Posts: 37
Liked: 4 times
Joined: Jul 21, 2016 5:03 pm
Full Name: Aaron B
Contact:

Synthetic Fulls Daily or Incrementals?

Post by adb98 » Apr 11, 2017 4:44 pm

Before I start, this was an EMC recommendation so go easy on me.

Here is the question...... With a Data Domain as the primary storage device, is it better to run a synthetic full daily instead of an incremental. With dedup and compression, we really are not loosing anything by doing a synthetic full I would think with compression and dedup. The restore times would be better as well as we wouldn't have to rehydrate multiple restore points.

During our meeting with our EMC tech rep today and he states that he doesn't see a reason why we just don't do a synthetic full each day and then run an active full once a month. My question was is this going to tax the DD and he stated that it should be able to handle it ( New DD4200) and that after the first synthetic full, the next one should be only rolling up the changes.

Would like to get thoughts from the community before I make any changes. Right now I am doing the standard incrementals with one synthetic full once a week and one active full once a month.

Mike Resseler
Veeam Software
Posts: 4825
Liked: 509 times
Joined: Feb 08, 2013 3:08 pm
Full Name: Mike Resseler
Location: Belgium
Contact:

Re: Synthetic Fulls Daily or Incrementals?

Post by Mike Resseler » Apr 12, 2017 6:13 am

Hi Aaron,

Do you have Boost? The creation of a synthetic full will be very slow when using a data domain without boost.

We have an entire best practices KB that I would advise you to go through: https://www.veeam.com/kb1956

Cheers
Mike

adb98
Enthusiast
Posts: 37
Liked: 4 times
Joined: Jul 21, 2016 5:03 pm
Full Name: Aaron B
Contact:

Re: Synthetic Fulls Daily or Incrementals?

Post by adb98 » Apr 12, 2017 7:31 pm

Yes we have DDBoost enabled and running. I have been thru that doc several times and we are pretty much following it to the T but it doesn't say wither synthetic fulls run each day would be a good or bad thing.

jmmarton
Veeam Software
Posts: 1237
Liked: 141 times
Joined: Nov 17, 2015 2:38 am
Full Name: Joe Marton
Location: Chicago, IL
Contact:

Re: Synthetic Fulls Daily or Incrementals?

Post by jmmarton » Apr 14, 2017 3:00 am

I'm not sure why he's saying you should do a synthetic full each day. Keep in mind that a synthetic full is still an incremental backup, but then a merge is performed. Obviously on Data Domain with Boost the synthetic full is very fast but in either case the incremental is performed. I don't see the benefits of rolling up changes daily vs once a week unless you're backing up all throughout the day. DD has a limit of 60 files in a chain (full + 59 incrementals) so sure if you're running a backup job say every 30 minutes and want to have a lot of restore points you'd need to do synthetic fulls daily.

Regardless, the data is all deduped, so the extra synthetics shouldn't take up any additional space. I just don't see any benefit. Did your EMC tech rep state *why* you should do daily synthetics? Unless he can state a specific tangible benefit, I wouldn't start reconfiguring backup jobs.

Joe

dellock6
Veeam Software
Posts: 5488
Liked: 1510 times
Joined: Jul 26, 2009 3:39 pm
Full Name: Luca Dell'Oca
Location: Varese, Italy
Contact:

Re: Synthetic Fulls Daily or Incrementals?

Post by dellock6 » Apr 17, 2017 8:06 am

We usually suggest to leverage the API integration we have (so DDBoost, Catalyst or ReFS Blockclone) on a weekly base, as this is an almost "free" activity but it also has some limits: you always have an incremental from production, but on target you split the incremental chain having a new full "for free". This said, I can imagine the EMC guys are suggesting this so that each day is not dependent on the previous days but still we are talking about a fully deduped system, where a block is pointed at in multiple times, so any corruption in those blocks can impact every backup pointing at it. Synthetic fulls can save you from metadata corruption, but not from data corruption, this one could be solved only with active fulls, or synthetic "NOT" using dedupe API. But this would obviously make the synthetic full a heavy IO operation. If you are ok running a synthetic daily, I don't see any problem doing so.
Luca Dell'Oca
EMEA Cloud Architect @ Veeam Software

@dellock6
http://www.virtualtothecore.com/en/
vExpert 2011-2012-2013-2014-2015-2016-2017-2018
Veeam VMCE #1

Post Reply

Who is online

Users browsing this forum: DonZoomik and 16 guests