- Your 3rd party application (if any and if hosts on same server as database) may or may not be certified to run on Solaris. Though usual practice in big organisation is not host application on a separate host than the database server host.
- Scripts running on database server needs to be tested on solaris. ksh and bash command behave differently on linux and solaris.
- Many organisations are considering or already shifting from SPARC on proprietary hardware to Linux on commodity hardware. Though companies could consider Solaris x86-64, but i have not seen many companies adopting it. If they go with Intel/AMD, its usually has linux/windows.

Lets hope Oracle officially certifies their database on OEL6/RHEL6. Though there is nothing technically stopping you from running Oracle right now on OEL6/RHEL6, apart from official certification.

Vishal

The fact of the matter is, none of us know why the certification has not happened yet and anything we say is just speculation, which is essentially just a conspiracy theory, regardless of what you want to call it.

Cheers

Tim…

markets targeted by Solaris/Sparc are not different to the ones targeted by Linux/X86_64. Most of the companies I know around here (European markets) , are Solaris/Sparc historical clients and initiated the move to Linux/X86_64 since a while (thanks to Nehalem/Linux successful combination) on business critical applications.
I’m talking about companies with more than a few thousands of servers (for a few hundreds of Oracle instances and Sybase dataservers) where standardization processes are based on vendors roadmaps and exhaustive solutions testings. In this case, the roadmap thing hurts.
There may be a revenge thing here, since RedHat have changed their errata policies making it more difficult for cloners (Oracle Enterprise Linux, Centos, …) to follow the original stream.
But still convinced that clients who have initiated the above move to hesitate and step back to the original Sparc/Solaris known solution due to the unclear RHEL/Oracle roadmap, and this is makes gains of market parts for Oracle.
I’m not in the conspiracy theory field here, but more on “fair war” acts.

Brem

Of all the conspiracy theories people have posted, the “annoying RH” one is the most plausible, but it would not surprise me if there were no sinister motives. Certifying the database on a new OS is a very big task. Remember, the database software is the same on Oracle’s Engineered Solutions and Appliances as what you run on your kit. That means certifying the database on RHEL6/OL6 means they probably have to give the green light to having OL6 on Exadata and ODA etc. That is a massive undertaking to be ready to support that sort of thing, especially when the UEK kernel in OL5.7 has all the performance features (and more) of the RHEL6.2 kernel already…

One conspiracy theory I’m mulling over is that 11g will never be certified on RHEL6/OL6. Instead, a move to RHEL6/OL6 could require an upgrade to database 12c. It will be out later in the year. It would be a neat way to try and force people into a quick database upgrade. Remember, this is just a bullsh*t conspiracy theory I’ve made up. There is no insider knowledge and no basis in fact. Even so, it seems more likely than a ploy to push T4 kit…

Cheers

Tim…

MOS Note 1350000.1 states that as of article on 21-Dec-2011, OEL6 and RHEL6 is not certified for Oracle database and client. Not sure why they are taking this long. Oracle themselves have released OL6.1 and OL6.2 since its base release of OL6.0. Surely Oracle has had plenty of time to test the database and client software on OL6/RHEL6.

Come on Oracle people are waiting for official certification so that they can start using Oracle on RHEL6 in production.

You can not use the USB image to install cell image in a VM. USB Image also tries to install all the firmware for ILOM, BIOS and other hardware which dont have in the VM. So trying to install cell in VM using the USB image will fail. Installing cell in a VM involves using a standard OEL5.5 VM server and carrying out manual steps to install the software. There are install scripts which can be run straight from the downloaded cell image tar file (skipping the firmware bits and cellos bits).

Vishal

[ERROR] get_node_type: Not a supported hardware model Virtualbox. dmidecode -s system-product-name returns invalid value

Ever came across this?

Thanks

Since i wrote this post, i have also installed exadata cell images in 3 separate VMs on my laptop to simulate a quarter rack. But installation process is not the same, its more involved. Like Gokhan Atil mentions in his blog post, i also can not post the detailed step-by-step process for legal reasons.

Installation process is drastically different than exadata hardware, but its possible to install the exadata software in a VM.

I have fixed the link.

Red Hat said they expected it last quarter, which there is still about 6 week of, so it could be any time now, assuming there are no delays.

I can’t claim to know Oracle’s motives, but certifying a platform is a pretty big deal. I guess Oracle want to be pretty darn sure before they do that and open the flood gates to support.

Cheers

Tim…

is Oracle playing the monopoly card and holding RHEL 6 back in the hopes that people go to OEL6?

Like I mentioned in point 8 above, password is not required for cell images from 11.2.2.3.2 onwards. If you download any older cell image version, for example for Exadata V2, instead of newer images which come in X2-2 or X2-8, then you will need password to extract contents of zip file, as it’s password protected. I happen to know the password though.

>> In newer cellimage from 11.2.2.3.2 onwards, zipfile is no longer password protected.

Vishal

Though oracle does have a Exadata simulator software which use internally for educational purposes but unfortunately they don’t give that our to customer.

Vishal

Could you please also let us know for what purpose Oracle uploaded Cellimage on E delivery site? Can this be used with oracle virtualbox or vmware.

Regards,

Good sharing especially about the connection between Automatic [Shared] Memory Management usage and decreased _small_table_threshold value after instance restart

We noticed the same behavior, which resulted in the significant increasing of some queries execution time, which used quick operations HASH JOIN / FULL TABLE SCAN of cached tables
After db restart this tables “suddenly” became BIG tables (blocks > _small_table_threshold*5) and we got a lot of “direct path read” events and physical reads (statistics) on EVERY execution

Igor

I was mainly referring to 10g method from Oracle white paper, but the method has changed significantly when compared to the deployment guide of CRS in 11gR2.

I will continue to look into it, but since you had done some work, I thought of checking with you first before I dive into it..

Thanks,

Yes, i have written scripts in past to clone 10gR2/11gR1 clusterware, ASM and RDBMS. I am just curious to know where did you see those scripts. Process between 10gR2/11gR1 and 11gR2 will be *lot* different. RAC Clusterware has changed drastically from 11gR2 onwards.

Following are some (and many more) of the changes in 11gR2 :-
- OCR and Voting disk are stored inside ASM now, instead of being on separate disks.
- SCAN listener is introduced, which was not there earlier.
- Server pool are introduced in 11gR2.
- CTSS (Cluster Time Synchronization Service) has been introduced in 11gR2.

I have tried cloning the 11gR2 clusterware yet. But may be i will do that one day, once i manage to find some time.

Regards,
Vishal

I have seen some of Oracle whitepapers regarding cloning with Oracle 10gR2 clusterware, and rdbms. I have seen some of your sample scripts regarding the same.

How is the clone process different in 11gR2 for clusterware, ASM and RDBMS? Have you had any experience with it? I have been following chapter 4 of Cluster Install and Deployment guide of 11gR2 but so far, I have not been able to clone the cluster.

Any suggestion/pointers would be greatly appreciated.
Thanks,

So if you are using RHEL, then yes, it’s good news. If you are using OL+UEK, the jump from OL5.x to OL6.x is not so drastic.

Cheers

Tim…

It will be cool when this is confirmed and the oracle-validated package is updated accordingly.

Cheers

Tim…

Thanks for stopping by.

Cheers,

Roel Hendriks

Have you tried event 10949 on an Exadata system? Serial direct path reads happened regardless of the event on Exadata when I last tried it. It did work as expected on non Exadata environments.

Cheers,
Tim

you are right, problem was solved from strace itself, but since the output was too long, i was focussing at the last section of strace for clue why it was hanging, rather than paying attention that it was calling the right RMAN command. I never expected a linux (non-oracle) rman command to be there. So i wrongly assumed that it was calling the right rman. Now i think back, strace always gices you the full path of command getting executed in the very first line of it.

Thanks for pointing this out to me.

No, i never worked in CMC

R u vishal worked in CMC(Mumbai) between 1999-2000
bye

Any update from Oracle on the drop user slowness.

Thanks
Andrew

I will have to check regarding extent size when i am in office again on Tuesday.

You have multiple PeopleSoft databases (schemas) in a single Oracle database. This is supported, but not recommended because of the additional complexity and lack of flexibility. And in your case it sounds like the shear size of the resultant data dictionary may also have a bearing. If you had only one PeopleSoft schema per database, then instead of refreshing schemas you can clone whole databases.

If you are thinking of doing any performance testing on these refreshed schemas, then bear in mind that you have repacked the data in the tables, and rebuilt the indexes.

Have the PeopleSoft tablespaces been created with a 1M uniform extent size (which is in the delivered DDL script)?

Mine was a simple spool script created to drop all objects (except LOB and package bodies) from dba_objects for a schema. Nothing fancy in my script.

Which script are you using for the drop objects as it sounds to be quicker than mine (which I will post tomorrow from work)

John

Example above was more directed towards DBAs. And dba would normally have access to server. Its using oracle directory object which is a directory on server.

table1 is just an example. It can be any table with a BLOB column. Lets say it has following definition.

CREATE TABLE table1
(COL1 BLOB);

I don’t want to user to Upload the file to the server from their client machine, and ask somebody else to load it from there.

I want the user to insert directly into the BLOB….

Is there any way to use this Procedure from the Client??

I tried, but it seems Oracle does not support client call to DBMS_LOB.

Congratulations with your site. It looks good it is interesting and wel organized.

The “Slow drop user ” and “Loading File into a Blob” I recognized….

Cheers,

Sander Westphal