Pages

Thursday, February 17, 2011

Solaris Cluster/kernel Patch Installation

Applicable for Solaris 8, 9 , 10.

Download the Cluster Pack from the sunsolve site. Download and keep it in the root directory. The file size may come up to 1 GB. Ensure ample space is in root file systems . ( ~ extra 1-2 GB for patches )

Precheck:

1) Take the logs prior to activity ( Suggested a day before the scheduled activity ). A script to take logs will soon be uploaded in here.
2) Check for hardware failures - Disk , CPU, Boards etc.
3) Check the console access availability - Console 3200 is known to hang sometimes , it is better to confirm prior the acitivy that you are able to access the box via console and run commands. If not, you may want to try rebooting or resetting the console server. Onsite engineer will be needed at the time of the activity otherwise.
4) Pull out the mirror disk . ( Best practise )


Activity:

1) Unmount all the file systems. ( Hash the NFS mounts in the vfstab before the activity )
#umountall
2) If mirror disk was not pulled, detach the plex from the mirror . Please note the mirror disk will not boot once the plex is detached. The disk needs to be un-encapsultated to make it bootable. How to un-encapsulate disk under veritas notes will be uploaded soon
#vxplex -g rootdg dis rootvol-02 swapvol-02 var-02
3) Bring the server to Single User mode
# init 0
#boot -S ( init S is not recommended )
4)cd /8_recommended { Path to the cluster patches )
5) ./install_cluster { Install cluster will install all the patches in the directory including the kernel patch. This normally takes 1-2 hours. Server reboot will be prompted at the end of patch installation after which the patch installation will be in effect )
6)Confirm the patches has been installed by runinng
#uname -a
7) Verify the logs
8) Release the box to DBA/Apps team


Issues Noted - Few Kernel patch installations have been observed not to take effect even though server is rebooted. Once the box was brought down to the boot prompt and rebooted with a reconfig option ( reboot --r ) the issue was found to be fixed.

No comments: