Update 7-May-2013: Almost 100 people filled in the survey and here are the result:
BCT survey results
I wrote a presentation on Oracle Block Change Tracking (BCT) internals more than 6 years ago. Back then, it was based on Oracle Database 10g Release 1. I have also written a paper in support of the presentation for Collaborate 2007 (You are coming to COLLABORATE 13 in April. Aren’t you?).
To the best of my knowledge, the core principles of block change tracking are still the same. In fact, Oracle recently asked me if they could publish this paper in the knowledge base on My Oracle Support. So now it’s there — ORACLE 10G BLOCK CHANGE TRACKING INSIDE OUT [ID 1528510.1]. I guess it means it’s still very relevant.
This reminded me that Fast Incremental Backups, enabled by Oracle BCT, is a very handy feature. I see lots of use cases that could leverage it, but for some reason don’t. Thus, I wanted to run a quick poll to see how the adoption of BCT is amongst the audience of this blog. If the poll embedded below doesn’t get displayed, you can follow this link.
As a reminder, Block Change Tracking enables Oracle database to keep track of changed fragments of datafiles between incremental backups. Using this log, it’s possible to identify the fragments of a datafile that changed after a particular incremental backup and then only read those fragments during the next incremental backup.
Block Change Tracking enabled us to deliver some innovative solutions to our customers. One unconventional use case is incrementally updated standby when you need to replicate a database that changes the same blocks over and over again and that, in doing so, generates so much redo that it’s not feasible to transfer that amount of redo to a remote site (like DR site) for replication. Transferring and applying incremental backups often requires much less bandwidth in these cases. Incrementally updated standby is also a way to replicate large data warehouses that are updated in NOLOGGING mode because standard physical standby doesn’t capture those NOLOGGING changes, so Data Guard for Data Warehouse is useless if you need to replicate NOLOGGING changes.
What are your use cases for block change tracking? Take the poll and feel free to provide more details in the comments section below on how you are using the Block Change Tracking feature or what prevents you from using it.
8 Comments. Leave new
Alex,
thank you again for your great paper!
Based on my observations there are no big new features on BCT even in 11.2.0.3:
select * from v$fixed_table where name like ‘%$KRC%’ order by object_id; still shows the same 10 tables with identical OBJECT_IDs. Only X$KRCSTAT has 5 new columns: HCI, PMRSLAVES, MEDIA_REC_START_SCN, NEWREDOSCN, OLDCKPTSCN – all Numbers. (in comparison to 10.2.0.5)
We are using BCT to decide ‘in case we already have enough good full backups within retention, is the amount of changes above a threshold, we do a full BU, otherwise an incremental BU is done’ on per datafile base.
Thanks for the update Martin.
[…] Pythian link White Paper MOS ID […]
[…] is Oracle Block Change Tracking today? World’s leading Oracle expert Alex Gorbachev […]
Alex,
Thanks for confirming the status of your paper. I’m currently working on enabling BCT on serveral hundred (11G) databases and your background work has been very useful.
A question on the current license situation with regard to RMAN backups.
MOS note 468576.1 states.
“You can only enable block change tracking at a physical standby database if a license for the Oracle Active Data Guard option is enabled”.
Is this still the case if you have BCT enabled on the standby but are backing up from the primary ?
Ron, I wouldn’t taking *my* licensing advice but based on my interpretation you need to license block change tracking whether you use fast incremental backups or not. If you don’t use Fast Incremental Backup on standby – why do you enable BCT there?
And for serious licensing advice I could recommend our good partner — Miro Consulting. They are awesome.
Alex,
Have you looked at the impact of BCT on a physical standby? When I tried turning on the BCT on my 8-node physical standby (trying to do incremental rman backups from there), it caused standby lag right away (lag 160 mins after 4 hours); after disabling BCT, the lag caught up quickly.
That’s a large system and will probably have to be investigated directly. Unfortunately, I haven’t really been on top of BCT for the last few years.