All 3 abstracts I sent were selected to be presented at UKOUG 2006! All of them on 14-AUG - tough day!
- Extending Oracle 10g Grid Control - New Metrics, Targets and Plugins
- Oracle 10g Change Tracking Inside Out
- Refreshing Test Databases Using Oracle Transportable Tablespaces and Veritas Storage Checkpoints
It’s going to be the first time I present to wide audience! Nervous tremor is getting stronger and stronger… %-0
I was looking on the grid computing syndrome and to my surprise Google Trends “Grid Computing” shows that hype around it was actually going down since 2004:
Interesting that the Oracle Grid Index shows continuous growth:
I can only think of some possible reasons for this deviation:
- People are moving from talking about it to really doing it.
- Oracle (well, Quocirca) publishes some biased or non-relevant results.
- “Senior IT influencers and decision makers” are overstating the figures.
- Google Trend graph is not relevant.
The second issue of Carnival of the Vanities for DBAs is out! Oracle July CPU topic was only on the third place whereas I expected it to be on the top… Still pretty close. 
I like very much that idea and hope that I’ll be able to contribute to this interesting project… since I am moving to Canada anyway. 
I also like very much that it’s not pure Oracle related - live and learn.
Actually, some bloggers were already regularly providing some links collections of the week/day/month (like Eddie’s links) but community driven project like that of Pythian looks very interesting.
Oracle histograms are often treated as devil of CBO as they cause bind variable peeking which is most of the times leads to quite nasty surprises when no one expects them. Common approach to avoid bind variable peeking is to remove histogram collection. However, it’s not true at all. Bind variable peeking can happen even without histogram and I am going to demonstrate with a simple test case. The only way I know so far is to use hidden parameter _optim_peek_user_binds (it’s still hidden even in 10g).
Originally, I found this issue in 9i and created a reproducible case. Later I also run in 10g and the result is the same. I will present here results as they are in 10.2.0.2 as there is more details in traces.
Read more…
Yesterday, I took a part in a meeting with one on Oracle product managers in security. Spoke about few technologies and Audit Vault looked quite attractive. On the other hand it’s nothing but a data warehouse that can load data from various sources including Oracle database, iAS and 3rd party tools. So if I can simplify it - it’s a bunch of Oracle BI tools integrated together and working on schema designed to analyze and mine audit data.
Another highlight was that Oracle has actually not much info about best practices in the area of security in HA environments. This is actually not as specialized as it might seemed to be. For example, what are implications of listener passwords - how is automatic startup/failover and etc. managed for password protected listeners. Or another one - how is Data Guard working with TDE (Transparent Data Encription). For example, automatic failover might require auto login to the wallet and/or wallet with no password. So it seems that there are many white spaces. Perhaps, time for a good paper?
James Foronda posted a nice note about ZFS. His primary topic was silent data corruptions but the presentation he mentioned is quite a good overview (though, smells a bit of marketing but who knows - maybe it’s all true?). In my previous post I mentioned that I am installing RAC on Solaris under VMware - I didn’t mention I use ASM for storage. I think the next step should be ZFS then!
Big news - I’ve got an exciting opportunity in Ottawa, Canada and finaly we decided to move. Quite a change as we were thinking for few years already to move to… Sydney! So quite an opposite place in the end with rather opposite climate. Once again I can see that my family is the best - they give me so much support here that no some are only dreaming about! Thanks to my wife Olga and my dear son Alex.
Alex Jr. is quite excited as he seems to be quite sure that bears are walking on the streets in Canada and he is going to ride bears overthere… also promise of snow, ice skating, skiing and etc. played the key role in his decision.
For me it was a very difficult decision as I really love my current company and the team I work in but this is an opportunity I couldn’t pass by. The interview (if you can call it that way) was quite a difference from what I usually would expect and I had an opportunity to chat with my future collegues and even customers! So I didn’t even start working but I already like the place. I couldn’t be mistaken - I know that feeling!
So I am looking forward to work with brilliant people at the new company on very interesting projects. I haven’t got much time recently to write here and probably going to have even less in the coming couple month.
My recent passion was installing two nodes RAC on Solaris on VMware - apparently this is quite a mission. I was thinking to combine my experience in a how-to document on installing Oracle 10g RAC on Solaris x86 with VMware. My current problem is performance - for some reason crsd.bin is spinning on CPU like crazy and I need to see why. So far shared storage issues are resolved, CRS is running happilly on two nodes, and Oracle installer is “hanging” in linking phase (well, all CPU is eaten by crsd.bin). I have already good start on the how-to document so I hope to have it ready some day.
Oracle CRS 10.1 didn’t provide any straightforward way to change to change a single property of a resource. Assume you need to change check interval of a VIP resource from standard 60 seconds to 10 seconds. The steps are below.
- Create .cap file with properties – “crs_stat -p ora.node1.vip >/tmp/ora.node1.vip.capâ€
- Edit /tmp/ora.node1.vip.cap to change line CHECK_INTERVAL=10
- Load .cap file back into CRS – “crs_register ora.node1.vip -dir /tmp -uâ€
Well not difficult but I was always wondering why not to provide a simple command-line option to change a single property. Our Oracle-God, probably, heard prays of DBA’s and added one improvement in 10.2 - crs_register has a new option -update to set some of attributes.
The format is -update [options] [-o options]
I am probably not as smart as developer and/or product manager who decided to split options in two groups - I failed to understand the reasoning behind… I guess as usual to make it tough to learn so that it keeps us in the job.
Anyway to change hosting nodes is “-update -h ” and to change check interval is “-update -o ci=10″. So instead of procedure described above you would use:
crs_register ora.node1.vip -update -o ci=10.
RTFM from more details.
“Are you missing your favorite scripts and indicators in Oracle Enterprise Manager Grid Control? Find out how easy it is to incorporate your own metrics into the Grid Control framework. Do you manage a multiple vendors’ environment and need to monitor many different components? Learn how to create your own completely new target types in Grid Control with rich set of availability and performance metrics.”
SELECT Journal published article “Extending Oracle Enterprise Manager 10g” in Quarter 2, 2006 issue.
Now you can download it on my web site here.
Interesting if Oracle will be able to push adoption of OEM so that it becomes the tool of masses. Perhaps, it’s a pricing model that prevents OEM from being popular.
I was reviewing abstracts for UKOUG conference 2006 and saw a lot of very good ones! I sent couple abstract myself but I am afraid that next to the famous speakers there mines would look faded. At least, I would understand if none of my presentations “make it to the finals”.
Well, Russians didn’t make it to the WM FIFA this year. :,(
Recent Comments