Best Practices for Oracle Linux for Production Systems

I found this needed a blog post because most of the customers I meet, just install their Redhat/Oracle Linux environment and start using it out of the box in production. I believe this list should be included in every post-installation procedure.

  • Hostname : Make sure it’s a FQDN. Especially when you connect with NFS to other systems. If your hostname is not FQDN, locks will not be freed on the NFS server when you reboot.
  • Support : If you have support, make sure you register your system with ULN.
  • Update : Update your system with yum or up2date to the latest version.
  • Hugepages : If you are running Oracle Databases, this is a must. Metalink note. 361468.1
  • Ipmitool : This allows for control over the hardware from inside the OS. Can be very usefull for Cluster setups or automated scripts to collect information.
  • Kexec : This allows the system to dump the kernel-memory to disk whenever a kernel panic occurs. Instead of rebooting or hanging, the system boots into a separate kernel with the task of dumping the memory to disk in the form of a vmcore file. This file can then later be analysed with the crash utility. Don’t forget to test it!!
  • magic sysrq key : This enables some key-strokes in the console to force a kernel to do all sorts of things ( show locks, reboot without FS corruption, … ). It is often used to dump a kernel stacktrace to /var/log/messages and reboot a system after soft hangs ( hangs on console with numlock flashing ). This is default enabled in OL5 but in OL6 you need to enable it manually. Also, make sure you know the keystrokes for when you need them.
  • Oswatcher BB : Monitoring tool of Oracle. Can show you if there were spikes just before or leading to the crash. Metalink note.301137.1
  • vncserver : allows for X11 environment over vnc. Faster then X11 over the net and allows you to continue where you left off when you lose your connection during an installation or configuration.
  • oratop : utility for near real-time monitoring of databases, RAC and Single Instance. Metalink note. 1500864.1.
  • dstat : allows you to view all of your system resources in real-time
  • Rlwrap: Saves you time 😉

If anyone is interested in how to perform some of these tasks, let me know and I’ll consider writing some blog entries about them. But most procedures can be found in the manual or official pages about it. Keep in mind that this list also applies to Oracle Engineered Systems.( ODA, EXADATA, … )

Leave a comment