This section applies only to systems running software release 2.0.5 or earlier.
The Oracle PCA Release 2.3 software automatically reconfigures the system software when changes in RAM are detected. You no longer need to execute the procedures in this section. If the instructions in this section were applied to a system with a previous software release, those changes are overwritten when the node reboots after the software upgrade.
Once the memory upgrade is installed you must apply specific configuration changes to make optimal use of the additional RAM and ensure correct provisioning functionality. The following changes are required:
reconfiguring the bootloader (GRUB) on the compute node with the memory upgrade
updating the PXE template for compute node (re-)provisioning
Reconfiguring GRUB on a Compute Node with Memory Extension
This configuration change is not persistent in the current Oracle PCA release. Reprovisioning the compute node results in the changes being overwritten, so you will be required to execute the same procedure afterwards.
Start the compute node after RAM installation.
Using SSH and an account with superuser privileges, log into the compute node.
Edit the file
/etc/grub.conf
and change thedom0_mem
value from the default 5888M to the correct amount of RAM installed. The entries are highlighted in the example below.NoteIf the total amount of RAM is now 512GB, set dom0 memory to 11264M.
If the total amount of RAM is now 768GB, set dom0 memory to 16512M.
root@ovcacn07r1 ~]# cat /etc/grub.conf # grub.conf generated by anaconda # # Note that you do not have to rerun grub after making changes to this file # NOTICE: You have a /boot partition. This means that # all kernel and initrd paths are relative to /boot/, eg. # root (hd0,0) # kernel /vmlinuz-version ro root=/dev/sda2 # initrd /initrd-version.img #boot=/dev/sda default=0 timeout=5 title Oracle VM Server (2.6.39-400.249.3.el5uek) root (hd0,0) kernel /xen.gz console=com1,vga com1=9600,8n1
dom0_mem=5888M
dom0_vcpus_pin dom0_max_vcpus=16 allowsuperpage crashkernel=256M@64M module /vmlinuz-2.6.39-400.249.3.el5uek ro root=UUID=f8b8e26c-532a-464f-ad0a-f535f8b70cea console=hvc0 console=tty0 console=xvc0 swiotlb=131072 module /initrd-2.6.39-400.249.3.el5uek.img title Oracle VM Server-ovs (xen-4.1.3 2.6.39-300.32.5.el5uek) root (hd0,0) kernel /xen.gzdom0_mem=5888M
dom0_vcpus_pin dom0_max_vcpus=16 allowsuperpage crashkernel=256M@64M module /vmlinuz-2.6.39-300.32.5.el5uek ro root=UUID=f8b8e26c-532a-464f-ad0a-f535f8b70cea console=tty0 console=xvc0 swiotlb=131072 module /initrd-2.6.39-300.32.5.el5uek.img title Oracle VM Server-ovs serial console (xen-4.1.3 2.6.39-300.32.5.el5uek) root (hd0,0) kernel /xen.gz console=com1,vga com1=9600,8n1dom0_mem=5888M
dom0_vcpus_pin dom0_max_vcpus=16 allowsuperpage crashkernel=256M@64M module /vmlinuz-2.6.39-300.32.5.el5uek ro root=UUID=f8b8e26c-532a-464f-ad0a-f535f8b70cea console=hvc0 console=tty0 console=xvc0 swiotlb=131072 module /initrd-2.6.39-300.32.5.el5uek.imgSave the changes to
grub.conf
and reboot the compute node to allow these changes to take effect.If you installed memory extension kits in multiple compute nodes, repeat these steps on each of those nodes.
Editing the PXE Template for Compute Node Provisioning
This configuration change is not persistent in the current Oracle PCA release. An update of the appliance controller software will overwrite the change, so you will be required to execute the same procedure afterwards.
Using SSH and an account with superuser privileges, log into the active management node.
Edit the file
pxelinux_cfg_OVM_template
located on the shared storage and make sure theAPPEND
line containsdom0_mem=5888M
as highlighted in the example below.[root@ovcamn05r1 ~]# cat /nfs/shared_storage/pxe_templates/pxelinux_cfg_OVM_template default OVCA-COMPUTE-OVM-1 LABEL OVCA-COMPUTE-OVM-1 KERNEL OVM/mboot.c32 APPEND OVM/xen.gz
dom0_mem=5888M
--- OVM/vmlinuz ks=nfs:XXXNFS_IPXXX:/export/MGMT_ROOT/kickstart/XXXKICKSTART_FILEXXX \ ksdevice=eth0 linksleep=50 --- OVM/initrd.imgThis value works for all supported RAM configurations. Compute node (re)provisioning will function as expected.