UKOUG 2012 – Day 2
Posted by John Hallas on December 4, 2012
A deep dive into Dataguard is a tough way to start the day off but Emre Baransel handled it well. I jotted down about 6 or 7 takeaways for later investigation. Much of the talk was about tuning the redo log apply and I must admit that I don’t consider we have many problems in that area across the estate but it is still worth reviewing. One topic that stuck me as looking at in more detail was the ability to recover a standby database from a primary backup. Chatting to a few people afterwards they had all heard of the capability to do that but had never tested it.
Maria Colgan is fanatastic. Nuff said. Very, very knowledgeable about the optimizer and a great speaker. She just seems so enthusiastic about what can be a very dry subject. Too many notes for me to mention but one worth highlighting is a tip when you wonder if your hint was used or even seen. Take a 10053 trace and search for ‘Dropping Hints’ which will be found right at the end. The used and error values will give you the answer to your question.
PS nice but firm shutting up of someone in the audience who liked the sound of his own voice too much.
Owen Ireland ran through an introduction to Goldengate which despite using GG was quite informative and helpful and had lots of pointers to useful resources. If looking at GG then I suggest you get the slides from the UKOUG library
Before lunch Larry Carpenter was talking about a new 12C feature which allows Data Guard to transfer to a remote (a long way away and therefore suffering from potential latency) standby. Far Sync allows a lightweight database to be sited locally but not that close that it will be affected with a primary site failover. A lightweight server with minimal memory, cpu and storage can be used (and it can host several Far Sync instances). Sync (max availability) transfer to there and that performs async (max performance to the remote standby database. It seems a good idea if the DR site is hundreds or thousands of miles away.
There will also be a DG administrator role in12C that will only be able to manage DG but not touch or see data.
My second Carlos Sierra talk of the conference was about the SQL Health Check script which is like a poor man’s SQLTEXPLAIN. However this script does not create any database objects so can be run on production systems without a change control. Well worth investigating, all you do is plug a sql_id in and then review the html page that is output.
Each year I try and attend a session about a subject that is new to me and this year I decided on Exalytics. The presenter was Robin Moffatt who gave a very accomplished overview of the toolset (which must be heaven for Oracle sales staff as it involved licenses for OBIEE, Times10, Exalytic hardware and Goldengate (optional)). Robin’s presentation was very professional and gave a really good overview, however the real value was where Robin talked about what worked and what didn’t and he obviously had plenty of experience of both aspects.
A good day spoilt by feeling ill during the late afternoon and evening.