Solved

Sealing a DDB - what happens to the old DDB?

  • 9 September 2021
  • 9 replies
  • 110 views

Badge +3

Hi Commvault-people.

 

I have a large partitioned DDB which has been writing to a Cloud-based library, and has been for some time. The DDB partitions are roughly 2 TB’s in size.

 

As is recommended when you have been writing to Cloud libraries, it should at some point be sealed, and I would like to go ahead. We are also on the cusp of the maximum threshold for Q&I times.

 

However, I need to make sure I have enough space for my DDB on the current volumes.

 

So the question is, what happens to the old DDB?

 

I am assuming that it will remain at 2 TB’s in size until there is a corresponding reference for the blocks in the new DDB, or the blocks eventually age and are therefore not required. Maybe that will take months. Quite probably. Whilst it ages out old blocks, will the old DDB reduce in size?

 

But what can I expect from the new DDB?

 

If I only have a 3 TB volume, and 2TB is taken up by the old DDB then I really only have 1 TB available.

 

if anyone has recently been through this scenario, it would be good to hear from you.

 

Thanks

 

MTG.

 

icon

Best answer by Damian Andre 9 September 2021, 19:27

View original

9 replies

Userlevel 6
Badge +14

Hi @MountainGoat 

You’re correct in your assumptions.

Sealing a DDB effectively closes it for new entries.

The sealed DDB will remain operational until all data is pruned.

As jobs for the sealed DDB become aged, records in the DDB are pruned. So the moment a DDB is sealed, data aging will begin to prune out records and the DDB will reduce in size.

It depends on the nature of the data and retention settings as to how much the DDB will reduce in size and over what timeframe.

For example, as deduped data is aged, secondary records are pruned from the DDB as these are simply references to the original unique data and primary DDB records. Once all the secondary records are removed for a chunk of unique data, the primary record is then also pruned along with the data on disk.

 

 

When a DDB is sealed, a new DDB is generated with no records, so the initial backup jobs that run will all be considered new, unique data.

Not only will the DDB grow in size, but your data written to disk/cloud will increase substantially as you will have a unique baseline data for the sealed DDB still remaining on storage as well as the new unique baseline data for the new DDB.

Effectively you need to ensure you have enough storage space to maintain 2 X baseline data.

Since the unique data for the sealed DDB will be the last data to be removed, you will need to wait for all secondary records to be pruned for chunks before primary records and unique data on disk is pruned.

It is expected that data aging for the sealed DDB will be pruning secondary, deduped records for some time and not the physical data on disk. You will need to expect disk space consumption to remain high for some time until you start to see primary records and unique data getting pruned to realise a reduction in storage disk space consumed.

 

Thanks,

Stuart

Badge +3

Thanks Stuart. Most of that does ring a bell but it’s been a while.

 

Fortunately, we don’t need to seal too often.

 

One more question if that’s ok?

 

I’m looking at moving these DDB’s onto better performing disk.

If I execute the move one partition at a time, will leaving “one-side” active allows processing to continue?

One the first move is complete, I will then release the second.

 

Thanks again.

Userlevel 6
Badge +14

Hi @MountainGoat 

I have been discussing with the team internally who have also run a couple of tests.

When you perform a DDB move, jobs associated with that DDB will be auto-suspended, so you won’t be able to continue running jobs, even moving 1 partition at a time. 

Thanks,

Stuart

 

Badge +5

Hey @Stuart Painter ,

 

themain question was will the new ddb in version 6 after sealing ? 

Userlevel 7
Badge +21

@SSchmidt , can you clarify the last question?

Userlevel 7
Badge +15

Hi @MountainGoat 

 

The sealed DDB will remain operational as it will be needed to reference dedupe data for restores and will remain open until all data is pruned.

 

Just a quick correction here - restores never reference the DDB - you could nuke your DDB from orbit and will be able to restore everything, but its true that sealed DDBs are still used for granular pruning - to ensure we can free up space as jobs meet retention and specific blocks become available. If the DDB were unavailable, we could not prune any data until ALL the associated jobs met retention, as we would not know which jobs reference which blocks.

Badge +3

Hi @MountainGoat 

 

The sealed DDB will remain operational as it will be needed to reference dedupe data for restores and will remain open until all data is pruned.

 

Just a quick correction here - restores never reference the DDB - you could nuke your DDB from orbit and will be able to restore everything, but its true that sealed DDBs are still used for granular pruning - to ensure we can free up space as jobs meet retention and specific blocks become available. If the DDB were unavailable, we could not prune any data until ALL the associated jobs met retention, as we would not know which jobs reference which blocks.


Thanks Damian.

 

Yep - Aware of DDB  not being required, but thanks for the additional info on pruning.

Userlevel 7
Badge +21

Glad everyone is helping!  Let me know if you have any questions we didn’t addressed!

Userlevel 6
Badge +14

@Damian Andre 

Appreciate the correction and extra guidance, I’ll edit my original post to take the restore piece out.

Thanks,

Stuart

Reply