installing Oracle 11g on gentoo

[keilb0ut]

If not nptl in global USE flags
1 add ntpl to global USE flags
2 re-emerge all ebuilds listed by equery hasuse nptl
3 re-emerge linux-headers (?)

If not installed public domain korn shell
1 emerge app-shells/pdksh
you can probably also create a softlink named ksh to sh, not tested

Start installation
1 simplify process by doing software only, create db later
2 in linking phase on linker error, add -lrt to $ORACLE_HOME/lib/sysliblist and run gcc-config -L, create softlink /usr/lib/libstdc++.so.6 to the one in this directory
3 press retry

After installation
remove created softlink /usr/lib/libstdc++.so.6

worked for me on 2.6.22-gentoo-r5

[alex.blackbit]

that sounds quite interesting. do you maybe want to put that knowledge into a shiny new ebuild?

[jonnevers]

[keilb0ut]

ditching Gentoo?

basically, there are two problems when i installed it on one of my Gentoo boxes. Firstly, i did not have nptl (new posix threading library) installed -and- on Gentoo you have to specifically link against nptl. i guess that on opensuse the new nptl is implicitly included in the standard linux libraries. Here, Gentoo gives you choice. The other issue with Oracle depending on ksh is a bit silly, Oracle should be able to divert to (or rather use the more standard) /bin/sh. Again, on Gentoo, the workaround is quite easy.

concerning an ebuild, i am not an ebuild author, so it would be my first attempt i shiver when it comes to licensing issues, and you would have to download the oracle 11g distribution by hand anyway. besides, the workaround is so trivial that i doubt it is worth the effort.

I have Oracle 11g on Gentoo running under load for weeks now (with async and direct IO). It works flawlessly.

[jonnevers]

[exitsfunnel]

Hey Guys,

Thanks to keilb0ut's helpful post I was able to get Oracle 11g (though not EM) running on Gentoo. Unfortunately, for unrelated reasons, I was forced to ditch that install of Gentoo and start all over with a fresh Gentoo install so I'm now about to take another crack at installing 11g. One question: how can I be sure that my system is already using ntpl? My impression is that since I have glibc-2.5 installed, that I am, but I'd like to be sure.

Also, I'm wondering if any of you who have got 11g up and running have also been able to get Enterprise Manager running. I've done a bit of googling and it seems that a lot of people installing on unsupported distros (eg, Gentoo) have had problems with EM.

Thanks.

-exits

[keilb0ut]

[dgaffuri]

Hi

First of all thanks for the post. I've managed to install 11g on my laptop but I've still a couple of problems:

• I've a repeaetd SIGSEGV message in $ORACLE_BASE/product/11.1.0/db_1/rdbms/log/alert_soto.log
  

  Code:

  Sun Oct 21 13:00:05 2007 Exception [type: SIGSEGV, Address not mapped to object] [ADDR:0x0] [PC:0xDBAF0A2, ioc_pin_shared_executable_object()+886]

  
  starting some minutes after instance starts. Related message in trc file is
  

  Code:

  Dump file /home/oracle/base/product/11.1.0/db_1/rdbms/log/soto_j000_1019.trc Oracle Database 11g Enterprise Edition Release 11.1.0.6.0 - Production With the Partitioning, OLAP, Data Mining and Real Application Testing options ORACLE_HOME = /home/oracle/base/product/11.1.0/db_1 System name:    Linux Node name:      sotomayor Release:        2.6.22-gentoo-r8 Version:        #1 Sat Sep 29 22:22:45 CEST 2007 Machine:        i686 Instance name: soto Redo thread mounted by this instance: 1 Oracle process number: 22 Unix process pid: 1019, image: oracle@sotomayor (J000) *** 2007-10-21 13:29:24.884 *** SESSION ID:(74.823) 2007-10-21 13:29:24.884 *** CLIENT ID:() 2007-10-21 13:29:24.884 *** SERVICE NAME:(SYS$USERS) 2007-10-21 13:29:24.884 *** MODULE NAME:() 2007-10-21 13:29:24.884 *** ACTION NAME:() 2007-10-21 13:29:24.884 peshmmap_Create_Memory_Map: Map_Length = 4096 Map_Protection = 7 Flags = 1 File_Offset = 0 mmap failed with error 1 error message:Operation not permitted *** 2007-10-21 13:29:24.885 Exception [type: SIGSEGV, Address not mapped to object] [ADDR:0x0] [PC:0xDBAF0A2, ioc_pin_shared_executable_object()+886] ssexhd: crashing the process... Shadow_Core_Dump = PARTIAL

  
  
• I get this message in $ORACLE_HOME/rdbms/log/alert_soto.log
  

  Code:

  Sun Oct 21 12:32:39 2007 ERROR: The process is unable to create the ADR schema in the diagnostic_dest directory ERROR: because of a disk issue or OS platform issue Sun Oct 21 12:32:39 2007 ERROR: Reverting back to using the user_dump_dest and background_dump_dest ERROR: as the location for the traces and logs

  
  and this message in $ORACLE_BASE/diag/rdbms/soto/soto/trace/alert_soto.log
  

  Code:

  Sun Oct 21 12:32:39 2007 ERROR: ORA-48178 encountered when checking if the process is able to ERROR: create the ADR schema in the specified ADR Base directory [/home/oracle/base] ERROR: The ORA-48178 error is caused by the ORA-48122 error. ORA-48122: error with opening the ADR block file [/home/oracle/base/diag/rdbms/soto/soto/metadata/ADR_INTERNAL.mif] [0] ORA-27040: file create error, unable to create file Linux Error: 22: Invalid argument ERROR: Check if the directory is readable and check if the OS version is supported for ADR. ERROR: The process will switch back to the pre-ADR method of tracing and logging.

  
  The directory is readable and writeable by the oracle user:
  

  Code:

  $ ls -ld $ORACLE_BASE/diag/rdbms/soto/soto/metadata/ drwxr-x--- 2 oracle oinstall 4096 Oct 21 12:41 /home/oracle/base/diag/rdbms/soto/soto/metadata/

  
  

Any idea?
BTW, I've also set

[keilb0ut]

The SIGSEGV disappeared when i increased the size of /dev/shm
shm /dev/shm tmpfs nodev,nosuid,noexec,size=2048M 0 0
this is for a sga=1G pga=250M system. i ran into trouble when i increased the size of the 11G sga from 500M to 1400M.
as a side effect the SIGSEGV has gone, which really seems like an Oracle bug not handling shortage on /dev/shm (or something related) too well in this case. sorry for not reporting this

The ADR issue (advanced diagnostic thing) is really a consequence of the sigsegv but is still a problem in itself. The line

Linux Error: 22: Invalid argument

is kind of telling, a Linux OS call is invoked with an invalid argument. This is either a platform issue or an Oracle bug. I do not have
this error but if have .mif files in the metadata directory, so that seems to be working. A platform subtlety seems therefore likely.

i guess that ora_diag_SID is throwing this ADR error and therefore generating this error, maybe you could do a strace (use strace -e file or strace -e desc to limit output) on it and combined with lsof find out on what file the error is, and what Oracle is doing with it. according to /usr/include/asm-generic/errno-base.h
#define EINVAL 22 /* Invalid argument */
look for EINVAL in the trace output, and work your way back

have fun!

[calvinrsmith]

I was able to install oracle 11g 64 bit on my 64 bit gentoo however I also needed to emerge libaio (asynchronous io) and create a gcc script that was found first in the path. A bonus is that my enterprise manager does work and can be started / stopped using emctl and my browser (seamonkey) works with it correctly.

here is my gcc script, this is only needed for installing oracle, once oracle is installed you can delete this script
#!/bin/bash

if [ "$1" = "-m32" ]; then
gcc32 -L/usr/lib32 $*
else
gcc64 $*
fi

[aru]

Hi,

I'm trying to install Oracle 11g on 64 bit Gentoo but it seems don't work...

error log:

[ibins]

Hi,

same problem here, but after renaming the directories $ORACLE_HOME/lib/stubs and $ORACLE_HOME/lib32/stubs to something else, linking went fine.
btw, I have found incompatible references to "libc_nonshared.a", referenced in the files lib32/stubs/libc.so and lib32/stubs/libpthread.so. "/usr/lib/libc_nonshared.a" should be 64 bit on Gentoo!

[s.hase]

This are my notices with which I managed to install Oracle 11g on Gentoo amd64. Maybe they are helpful to someone. Don't forget to adjust the paths. This is not a complete HowTo, only the most important stuff.

Create user:

[ibins]

[slizo]

[s.hase]

[goldgriff]

I am attempting to install 11.2.0 and running into the same type of linker issues. Various different errors that are related to the libc.so and libpthread.so noted above. Anyone have a solution?

[goldgriff]

Modify $ORACLE_HOME/lib/stubs/libc.so

[RAPHEAD]

I have everything on the latest stable level for AMD64.
The "ltr" trick works for me although I still get linking errors during install.
Without the trick I have an unusable install.

[danky]

I followed the guide here: http://www.fuzzy.cz/en/articles/installing-oracle-11g-r2-on-gentoo/ to install oracle 11g r2 on my gentoo x86 machine. There were a few differences in the install screen process, and the error popup that was warned about did show up and was corrected by the fix mentioned. However I did have another error popup during the linking phase that I could not find a solution to and ended up just clicking continue past it. It was something like:

"Error in invoking target install of makefile / opt/oracle/product/11.2.0/ctx/lib/ins_ctx.mk"

Other than that the install finished successfully.

[masse]