Revisited: Installing OEL 5.4 in preparation for Oracle database 11gR2

Linux, Oracle

My OEL 5.3 server with Oracle hasn’t been behaving quite right lately, so all in all, the best choice at this time was to reinstall. At the same time, I wanted to redesign the disk / partition layout.

So here’s a quick recap of this installation.

First, the OS.

  1. Downloaded OEL 5.4 64-bit CD’s 1,2,3,4 and 6 from edelivery.oracle.com
  2. (This server has no DVD player, and previous installations taught me that CD 5 wasn’t necessary for me.)

  3. Booted from first disk, deleted all existing volume groups and selected two (out of six) disks to be available for the OS install.
  4. (Which means Linux will put them together in one Volume Group.)

  5. Disabled SELinux for the time being.
  6. (To prevent some problems when configuring Oracle later, specifically “cannot restore segment prot after reloc” error.)

  7. Created the oracle OS user during OS installation.

That went well enough, so now for some preparations.

  • In one session: Transfer Oracle software to the new server: rsync --progress linux.x64_11gR2_database_?of2.zip oracle@newserver:/src/
  • (Including mkdir /src and making oracle the owner of it.)

  • In a parallell session: Set up Oracle public YUM server, courtesy of
  • http://public-yum.oracle.com/

  • Install required packages: yum install oracle-validated
  • Create OS groups for oracle: usermod -g oinstall -G dba oracle
  • Make sure there is a line in the hosts file for the host itself, so the Oracle installer can map the hostname to an IP.

Create LVM volumes for the remaining disks; here are the commands used:

pvcreate /dev/sdc1 /dev/sdd1
pvcreate /dev/sde1 /dev/sdf1
vgcreate OracleVol01 /dev/sde1 /dev/sdd1
vgcreate OracleVol02 /dev/sdc1 /dev/sdf1
lvcreate -L 100G -n u01 OracleVol01
lvcreate -L 100G -n u02 OracleVol01
lvdisplay
mke2fs -j /dev/OracleVol01/u01
mke2fs -j /dev/OracleVol01/u02
e2label
e2label /dev/OracleVol01/u01 u01
e2label /dev/OracleVol01/u02 u02
vi /etc/fstab
mkdir /u01
mkdir /u02
mount /u01
mount /u02
df -h

And then, install Oracle software, which is another story :)

sh /src/database/runInstaller

Fixing Linux problem: “passwd: Authentication token manipulation error”

Linux, Oracle

I had a very irritating problem on one of my OEL 5 virtual machines, actually on my main Oracle 11g database test server, which is the last stop before production.

Googling did not solve this problem for me, since this error can apparently be caused by several things, so it was unresolved for many weeks, and I considered just creating another virtual machine instead, and reinstalling Oracle.

But this morning, I easily solved the problem, just by reading a man page a little more closely, so I thought I’d share my experience :)

Problem:
For some unknown reason, I could suddenly not change the password of my oracle user, neither as root or as the user itself, because using the command passwd immediately gave me the message “passwd: Authentication token manipulation error”. I could log in as the user, though, for example by doing su – oracle, or by ssh-ing in with public keys set.

Google thought this had to do with a lot of file access or pam settings issues that I knew were not relevant to my problem, partly because I always received this error before I got to write any new password for the user at all.

Solution (for me):
passwd (when run from the root account) also has the option -d, to delete (set to null), a user’s passwd. It occurred to me that this might work, since it might not involve any checks against the existing, somehow corrupt, passwd. And it did. So, for me, the 10 second solution was – as root:


passwd -d oracle
passwd oracle

Hope it might be helpful to others!