PowerFlex: 14G SO nodes idle settings revert to default after service upgrades

• The default settings put the CPU to sleep when SO nodes are idle to save power. In that state when we issue IOs it takes the CPU time to wake up and it impacts the response time of the IOs. So customers may want to prevent the CPU from going to sleep.
• This can be done by changing the grub settings to processor.max_cstate=1 intel_idle.max_cstate=0 intel_pstate=disable
• This increases the performance by making sure that the CPU does not go to sleep.
• These settings revert back to e default whenever a service upgrade is performed.

PowerFlex Manager only currently supports the default settings.

1. Open the GRUB template for editing:

   vim /etc/default/grub

2. Find the GRUB_CMDLINE_LINUX configuration option and append the following to the line:

   processor.max_cstate=1 intel_idle.max_cstate=0 intel_pstate=disable

   Example:

   GRUB_CMDLINE_LINUX="crashkernel=auto rd.lvm.lv=rhel/root rd.lvm.lv=rhel/swap rhgb processor.max_cstate=1 intel_idle.max_cstate=0 intel_pstate=disable"

3. Compile the new GRUB: See OS documentation. You could try to search for grub.cfg to get the location.

   IF EFI (replace <os> with Red Hat or centos)

   grub2-mkconfig -o /boot/efi/EFI/<os>/grub.cfg

   IF BIOS

   grub2-mkconfig -o /boot/grub2/grub.cfg

4. Stop and then disable the tuned:

   systemctl stop tuned
   systemctl disable tuned

5. Place the host into PowerFlex Maintenance Mode (IMM or PMM). If this is a Primary MDM, switchover to another MDM

6. Reboot

7. After the Host is up, check if the settings were applied successfully

   # cat /proc/cmdline

8. Edit the host out of IMM or PMM

9. Continue the same process for the other Storage only nodes.