When the tuned daemon dies or is killed, some kernel tunings revert back to their defaults
Issue
-
When we enable a tuned profile we can see the effect; however, we're concerned that a failure of the
tuneddaemon would lead to our server reverting back to a default (non-optimized) configuration.
In the following example, note that while tuned is active, all of the processors are spending NO time in deep sleep state (C6-Ias reported bycpupower).[root]# tuned-adm active Current active profile: network-latency [root]# cpupower monitor | awk -F\| '{print $(NF-3), $NF}' Idle_Stats C1-I C6-I 99.99 0.00 99.99 0.00 57.34 0.00 99.99 0.00 99.86 0.00 99.96 0.00 99.99 0.00 100.00 0.00When tuned is killed (simulating a failure), the CPUs revert back to waiting in deep sleep.
[root]# ps -fC tuned UID PID PPID C STIME TTY TIME CMD root 19754 1 0 11:37 ? 00:00:00 /usr/bin/python -Es /usr/sbin/tuned -l -P [root]# killall tuned [root]# ps -fC tuned UID PID PPID C STIME TTY TIME CMD [root]# tuned-adm active Cannot talk to Tuned daemon via DBus. It seems that tuned daemon is not running, preset profile is not activated. Preset profile: network-latency [root]# cpupower monitor | awk -F\| '{print $(NF-3), $NF}' Idle_Stats C1-I C6-I 10.27 89.72 99.99 0.00 0.70 99.28 22.91 77.08 0.40 99.47 0.00 99.95 77.92 22.07 0.70 99.30What can we do about this?
We're considering staying with the original method of simply using/etc/sysctl*and the kernel command-line
Environment
- Red Hat Enterprise Linux 7
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
