Having good friends in Oracle is key. And honestly one of the best things in my role. Philippe Fierens, product manager for Oracle’s Fleet Patching and Provisioning mailed me yesterday night sharing that there is an interesting change coming with Database Release Update 19.27: Small Pages not allowed for SGA from 19.27 and newer on Exadata.

Photo by George Bonev on Unsplash
What does this mean?
As you can read in MOS Note: 3081878.1 – Starting DBRU 19.27 Small Pages Are Not Allowed for RDBMS SGA In Exadata, from Release Update 19.27 onward (this is the April 2025 RU), you can’t start a database on Exadata anymore with a small pages SGA. This means that for instance memory_target (luckily!) will not work anymore.
The error you will see during startup is: ORA-27138: unable to allocate large pages with current parameter setting. As a consequence, the database will not start.
The same behavior will be seen from Oracle Database Release Update 23.8.0 onward as well.
Important addition since some people asked:
This will be enforced on Exadata only. It does not affect other platforms or environments.
In addition, the MOS note has been updated recently on April 18 to clarify settings and add more information.
What do you have to do?
- Remove MEMORY_TARGET
- Set SGA_TARGET
- Set USE_LARGE_PAGES=ONLY
.
Further Links and Information
- MOS Note: 3081878.1 – Starting DBRU 19.27 Small Pages Are Not Allowed for RDBMS SGA In Exadata
- SGA_TARGET
- USE_LARGE_PAGES
–Mike
Thanks for the heads up Mike.
Is it expected to be enforced also for non-exadata customers in the near future ?
Hi Ansel,
the patch is generic but I wait for reply from the developer.
I will update the blog post as soon as I have more information.
Cheers,
Mike
So is this a hint about what’s to come for non-exadata installations? i.e. that small pages will not be allowed anymore.
No – and luckily the MOS note has been precised. I updated the blog post as well.
Cheers,
Mike
Hi Mike!
What about the use_large_pages=exadata_only? Do you have an idea if this will work soon as designed? Would be a great step forward in handling huge pages!
Many thanks
Christian
Hi Christian,
based on what I have been told, the option “exadata_only” has been removed.
Cheers,
Mike
Thanks for reminding me Mike. Next time I’ll write it down 😉
😉
Cheers Christian!
Mike
Hi Mike,
Thanks for the article. Do you have any info when the bug described in 3064205.1 will be completely fixed?
It’s so unprofessional to introduce such bug and how unprofessionally it’s being fixed:(
Hmmm … hi Nariman,
did you mention the correct MOS note?
The number you are referring to has nothing to do with my article since it is:
Upgrade From 19.23 to 19.24 Receiving ORA-4061/ORA-4065/ORA-04021 On SYS.DBMS_RANDOM During Datapatch (Doc ID 3064205.1)
So, before accusing me/us to be unprofessional, please tell me more about your actual question. Then I may be able to either help you or direct you to the right person who can help.
Thanks,
Mike
Why does Oracle default to DONT START MY DATABASE when only half the memory on a server is consumed (memory is misconfigured)? Its one thing to warn, or scan the databases with exachk and warn. Entirely different thing to not allow a database THAT WAS RUNNING to not be able to start without a freaking SERVER reboot on a system running hundreds of databases. These patches should definitely not randomly hose your database. A patch successfully installed in the dev environment may not work in production. FUN.
Well, not my decision to handle it this way. But I guess there may be reasons why development enforced it now.
I can’t tell you the background about this unfortunately.
Cheers,
Mike