Wednesday started off by waking up early with a truly thumping headache, which was caused by drinking a very moderate quantity of red wine. A totally undeserved hangover The headache seemed to remain for most of the morning. I’m not sure rman is the best topic to mix with a hangover. I was however cheered up by bumping into Chen/Gwen Shapria.
My first session was by Asif Momem. Which covered a wide range of rman recovery scenarios. Asif Obviously has a really good grasp of backup and recovery concepts and procedures, but really the presentation would have been better with actual demos rather than just showing the various scenarios by screenshot. It’s a shame, as it was pretty comprehensive set of scenarios, but really sometimes less is more, and there is something much more instructive about seeing the demo occur in real time. He has a good idea about using a usb pen drive to help simulate failures (easily removable).
Next up was another rman sessions, this time by Timothy Chien, who is the rman prinicpal product manager, Greg Green a dba from a well known pedlar of coffee. This was essentially a how do you make rman go faster session, and it was a good discussion of the various bottlenecks involved and how some have trade offs, e.g. multiplexing the backup so that different datafiles are written next to each other in the backupset, this may well be faster to write, but when coming from tape having to possibly skip datafiles you are not interested in recovering is going to lengthen the recovery time. Some emphasis on tuning buffer sizes, but Oracle saying 11gR2 and ASM the buffer sizes are already automatically set correctly. It was interesting hearing a bit about the coffee pedlar’s warehouse environment, in particular they were saying a huge variance in their sunday full backup going from 15hours up to 30 hours. Seems like a spot of buffer tuning helped them by about 50%.
Last presentation of the day for me was a tuning Oracle RAC one, by Michael Zoll. There were a few nuggets, in particular Michael was emphasising that you need LMS to be able to get on cpu when it needs to, and therefore it runs in Real Time mode. However there are a few bugs with this, particularly 11gR2. Mentioned something about ocr/voting disks on raw being deprecated, but I hope he is talking new installs rather than upgrades.
By this time it was 2:00pm and i missed lunch, I took a bit of downtime in the afternoon and managed to fairly humiliate myself in a Chinese restaurant, at least the staff could spot an idiot when they saw one. Don’t mention Dim Sum. Still glad I skipped the Larry keynote, which I heard overran by quite some time, and then it was the Bloggers Meetup organised by Pythian. Alex Gorbachev had a rather resplendent bue shirt on to begin with. It was a fantastic event, and it was really great to meet some people I have been following from a distance for a number of years (no, not in a stalker kind of way): Christian Bilien, Tim Hall, who by the way, is far younger looking than perhaps his online presence would have you believe. Tim was demonstrating his new iPad and somewhat bemoaning the cut ‘n’ paste features. Yes the pdf’s of the Oracle manuals did look good. I Also found out that Paul Vallee would have made a terrible priest, and I narrowly escaped offending an American. It was also great talking to John Piwowar, and learning the correct pronunciation of his second name (note: it’s not like Pee). Definitely one person I was keen to seek out was Chet Justice, aka: Oraclenerd, who I think is quite an inspiration. It was a great meetup, really enjoyable.
And then off to the Appreciation Event with the irrepressible Martin Nash.