1. INFINIDAT Support
  2. Knowledge Base
  3. Host Readiness
Using the "noop" Linux I/O scheduler for better performance

Follow

Test Description

The tools checks that all NFINIDAT scsi-disk devices uses the noop I/O scheduler.

Linux provides a mechanism over sysfs to control the current I/O scheduler for every block devices, but no mechanism to make this change persistent.

To make it stick for all NFINIDAT devices, current and future ones, and across reboots, a udev rule must be added.

 

 

This test case fails if:

  • The tool checks our udev rule exists for setting the noop I/O scheduler for NFINIDAT sd devices
  • The tool checks that active I/O scheduler for all NFINIDAT sd devices is noop

 

This test case passes if:

All NFINIDAT sd devices uses the noop I/O scheduler, and the udev rule exists

 

How to fix this test

Automatic fix

The tool follows the instructions mentioned below in the manual fix

Manual fix

To install the udev rule:

  • on legacy udev systems, such as redhat-6, suse-11 and their equivalents:rite the following contents to /etc/udev/rules.d/41-infinidat-io-scheduler.rules:

    ACTION=="add|change", KERNEL=="sd[a-z]*", SYSFS{vendor}=="NFINIDAT", RUN+="/bin/sh -c 'echo noop > /sys$DEVPATH/queue/scheduler'"

     

  • on systemd-based systems, such as redhat-7, ubuntu-14.04 and their equivalents:rite the following contents to /lib/udev/rules.d/99-infinidat-io-scheduler.rules:

    ACTION=="add|change", KERNEL=="sd[a-z]*", ENV{ID_VENDOR}=="NFINIDAT" RUN+="/bin/sh -c 'echo noop > /sys$env{DEVPATH}/queue/scheduler'"

To change the active I/O scheduler for sd devices, run 

echo noop > /sys/block/sdX/queue/scheduler

for every NIFNIDAT device.

 

The license could not be verified: License Certificate has expired!

 


Comments