I finally found my top Orace 12c-19c database feature !

Every Oracle database version comes with tons of new features (or renamed features to kind of re release a feature which was not really well implemented in a previous version :)) which are advertised a lot but when you think back about these features, which ones were really the top new feature of each version ?

I can easily name what I consider (this is indeed very subjective) to be the best feature of each version up to 11g but honnestly, for the 12c family (12, 18 and 19), I had nothing in mind (also may be because I have been doing less database work for years now). Let's start by my top feature for each version:

• Oracle 7: CBO -- I haven't worked that much with Oracle 7 though but well, CBO is a huge feature


• Oracle 8: RMAN -- We finally had something more integrated and more efficient than BEGIN/END BACKUP which was kind of a hassle. I also remind funny stories when interviewing to recruit people: "Me: do you know RMAN ?", "Candidate: Hermann Maier ? yes he is very good at ski racing", "Me: OK, we'll call you" :D


• Oracle 8i: if 8i is a different version than 8 (some were saying that at that time), I would say Java in the database ! haha joking obviously; not sure this was the best idea ever as we still cannot really patch it online 20 years later and it is still full of vulnerabilities we have to fix every single month :) I would then say Partitionning -- 8 had partitionning but it was very basic, it started to be really usable in 8i; if 8i is not a different version than 8 then I go with RMAN for 8/8i


• Oracle 9i: dbms_metadata.get_ddl -- how awesome it was when get_ddl was released ! what a pain it was to generate a DDL in the previous versions ! query tons of different data dict views, doing export and clean the DDL from the export file, using some graphical tools, ... there was no easy and 100% reliable way, get_ddl was clearly a relief, long live get_ddl ! :)


• Oracle 10g: AWR -- no need to argue here, there was clearly a before and an after in term of performance investigations and tuning thanks to AWR


• Oracle 11g: Snapshot Standby -- we could already do this in 10g with restore points, I remember I had scripted this in the past but Snapshot Standbys made it easy (and official !) to open read write a copy of a production to developper and resync it at night, excellent feature; I could also have said Exadata ! but at the time Exadata was released, it was still kind of confidential, it took time for people to use / trust Exadata so for me, Snapshot Standby stays the top 11g feature

Then 12c came and I couldn't find anything really appealling to me. I was thinking about the online datafiles move which is indeed a very cool feature but Oracle has mainly addressed these kind of datafiles placement issues in previous versions with OMF, db_file_create_dest, etc ... so even if this is a very cool feature, it would have had a bigger impact if implemeted in 8 than in 12. One would say CDB/PDB; it is indeed the major architecture change but I personnaly don't really like it and making it default makes everything more complicated for 99% for may be 1% who would really need this feature so for me, it is not really the best 12c family feature. Also, it is not totally integrated with CRS which is why I cannot show the PDBs in rac-status.sh. It should be coming with GI 21c (as GI 20c has not been released) -- we'll see.

And suddenly, with 19c, came my top 12c family feature:

• Services can now automatically failback to the preferred node using -failback yes !!

Oh boy, I have always been waiting for this feature ! For everyone who have done patching / upgrades (everyone ?), you know how messy it is with the services: they restart on the available nodes when you patch the preferred node, application teams want to run on a specific node because they application is still not RAC-friendly or whatever reason, etc... You can easily know what services have moved during a maintenance using rac-status but when it comes to rebalance everything to the preferred node(s), this is another story. Indeed, CRS does not show the preferred / available nodes information with crsctl and you then have to svrctl config service to get this information knowing that you need to use the srvctl command from the database home version (and not CRS !) for this as a wrong srvctl version used for this would fail, .. in short, it is a big mess for something which should be easy in my opinion.

But this was before -failback yes feature ! you just have to set it up on a service like this:

srvctl config service -d DB_NAME -s SERVICE_NAME -failback yes

Then at next restart of the server, GI or instance, the service will automatically been rebalanced to the preferred node -- gracefully, without disconnecting anyone -- very cool, right ?!

To help setting that up, I have then wrote 2 scripts to be able to easily take advantage of this feature:

• svc-show-config.sh to show the current services configuration
• svc-set-failback-yes.sh to to set the failback option to your services

Enjoy !