UKOUG conference day 2 opened up with Tom Kyte talking about 5 PL/SQL things you probably didn’t know. It did have 12c in the title but I think 4 out of the 5 were pre 12C. Implicit conversions are EVIL is my main take-away and he made me aware of the power of the PL/SQL warning framework to highlight any implicit conversions and other no-nos. Another important warning was when you code a WHEN OTHERS statement but do not follow it with a RAISE statement. Overall, an interesting discussion which I enjoyed despite only having ever written about 200 lines of code in my whole life.
Larry Carpenter then talked about new Active Dataguard new features in 12C. To be honest I could not see a use for most of them at my current workplace but they were obviously delivered to meet a demand. One feature was DataGuard Fast Sync and that seemed to be available when using normal (inactive??) DG on 12c so it might be something to look at. Larry also mentioned that Oracle would be announcing g a database backup appliance in the near future.
Two roundtables followed, both of which were a bit like the curate’s egg – good in parts.
The first was around Oracle’s OEM offering. I was hoping it would be a meeting where the audience shared some good things they had discovered about OEM (around 60% of the 30 strong audience were on 12c OEM). My humble offerings might have documentation which provides as list of all the views on the sysman tables used to manage OEM.
However it really turned out to be of benefit to the two Oracle product managers in attendance as they just soaked up a list of issues from the audience – licensing difficulties and poor documentation mostly. There was one good tip on the help capability. If you type in backup or whatever else you want help on in the field on the home page where you put the target name in that will take you to help functionality that quite often has a direct link to the right page to get started.
One comment from the Oracle people was that they had an OEM channel on YouTube with 71 videos on it. All very good but anybody running a significant estate from OEM is probably a corporate and they are likely to have YouTube blocked at work. Nothing to stop you looking at home of course but it did seem a bit of a contradiction to me.
Then a three way round table on virtualisation that involved Microsoft, VMware and Oracle. The surprising thing was that they seemed to agree that the real benefit, above cost was agility. In fact they all seemed to agree on everything, especially the fact that there was no reason why the whole enterprise could not be virtualised. I raised the point that why would I risk putting a Tier 1 Oracle RAC database on virtual when Oracle state that in the event of a problem they might insist on getting it on physical before they could help further. I think I was seen as a Luddite with my head buried in the sand. However, given that it could take days to provision a physical RAC environment and get a copy database deployed how could I justify that to management. Microsoft and VMware said they would have their own level 2 /3 support who would help out but that is hardly the same as working with the main supplier.
Lunch was spent talking to a friend around disk latencies, , UKOUG and Apex. An enjoyable and productive 45 minutes.
In the afternoon the best talk was by Pythian’s very enthusiastic Marc Fielding on Private cloud. A few thoughts struck me on the container database and pluggable databases. Firstly it seems to be where SQLServer has been for years. The idea that you get try one pluggable database without licensing is bizarre because most of the things you would wish to test require more than one PBD. My two overall questions would be around memory management and security. With a single SGA that is not resource managed how does one stop one PBD filling the buffer cache after running FTS or producing a fragmented library cache with multiple parsing of sql – if you flush the shared pool does it do it across all the databases in the container. My main concern would be around security, a user on one PBD would be able to see the sql-text and v$session views of another PBD and therefore see data. I have not investigated CPD or PDBs at all so I don’t have the answers but I will be looking at it shortly.