On 7 Aug 2004 15:33:43 -0700, David Fitzjarrell <fitzjarrell_at_cox.net>
wrote:
> David Freeman <$d$a$v$i$d$@nospam.davidfreeman.net> wrote in message
> news:<pan.2004.08.07.17.30.25.953782_at_nospam.davidfreeman.net>...
>> On Mon, 31 May 2004 19:36:33 -0700, andi wrote:
>>
>> > Hi All,
>> > I installed Oracle 10g (10.1.0.2) on Fedora Core 2 (kernel 2.6.6.x),
>> > RAM: 512 MB swap 1GB.
>> > followed the instructions at www.dizwell.com, the software
>> > installation went smoothly, but then when I tried to create a database
>> > using dbca, there's an error message: ORA-27125 unable to create share
>> > memory segment.
>> > Tried to export DISABLE_HUGETLBFS=1 as suggested at OTN linux forum
>> > with no luck.
>> >
>> > Any other way to solve the problem?
>> >
>> > TIA,
>> > andi
>>
>> I successfully installed the Oracle 10g software on Fedora Core 2 today
>> and found the same error, which persisted even after I followed
>> the dizwell.com instructions regarding kernel memory, checked
>> gcc-related, rpm-installed executables, and all the other
>> preinstall steps. I got the error, then following supplementary
>> instructions exported the environment variable DISABLE_HUGETLBFS=1 as
>> you
>> did.
>>
>> Renaming the $ORACLE_HOME/bin/oracle executable to oracle.bin and then
>> creating a script that invokes it and ensures that this variable is
>> utilized:
>> =======================================================
>> This from http://www.orafaq.com/msgboard/linux/messages/568.htm :
>> =======================================================
>> cd $ORACLE_HOME/bin
>>
>> mv oracle oracle.bin
>>
>> cat >oracle <<"EOF"
>> #!/bin/bash
>>
>> export DISABLE_HUGETLBFS=1
>> exec $ORACLE_HOME/bin/oracle.bin $@
>> EOF
>>
>> chmod +x oracle
>>
>> =========================================================
>>
>> I then ran dbca and it worked to create a generic data warehousing
>> database, just moments after it had *not* worked when the
>> DISABLE_HUGETLBFS variable had been set
>> = 1. As with other first-time, one-off installations of the Oracle
>> software on Linux, I am somewhat unsure that this is the turnkey measure
>> for getting a database created just because it is the last thing I did.
>>
>> You may get errors when running dbca if it stopped somewhere around 48%.
>>
>> If you've just installed the software and you switched out your
>> /etc/redhat-release file out instead of editing
>> Disk1/install/oraparam.ini
>> file, don't forget to switch it back so that your OS knows its name!
>>
>> I hope this sets you right; please repost to the newsgroup to confirm.
>>
>> Sincerely looking for a job right now,
>>
>>
>> David Freeman
>> http://www.davidfreeman.net/DavidFreemanResume.txtOn Tue, 01 Jun 2004
>> 21:39:34 +0200, Igor Racic wrote:
>>
>
> Renaming the oracle executable to oracle.bin and writing a wrapper
> script was unnecessary since you could easily place the 'export
> DISABLE_HUGETLBFS=1' code in the .profile or .bash_profile for the
> oracle Linux user account, thus setting it without the workaround you
> coded.
>
> David Fitzjarrell
Interesting. I was about to post back to the OP along the lines of David
Freeman's workaround. And then I was going to post back to David Freeman
to assure him that the wrapper script was indeed the one thing he altered
which made it all work. Now the other David posts that what the wrapper
script does is just as easily done with an 'export' command somewhere...
despite the fact that exporting that environment variable in precisely the
same way David suggests does NOT make Oracle 10g work.... whereas the
wrapper script DOES.
It is rather odd, because I can't see the script does much that the
inclusion of 'export DISABLE...' in .bashrc or somewhere similar wouldn't
do. But they aren't equivalent, in my experience, and the wrapper script
is indeed needed.
Some Linux guru will no dount explain why they aren't actually achieving
the same thing, I am sure!
Regards
HJR
Received on Sat Aug 07 2004 - 22:23:29 CDT
