tuned
tuned copied to clipboard
redhat-performance
Set elevator for disk fails
I'm trying to configure a different elevator for different disks, I've setup the following Ansible template that and seems to works well.
[main]
summary=Customized low-power usage profile
include=powersave
[bootloader]
cmdline=pcie_aspm=force
[sysctl]
net.ipv4.tcp_rfc1337 = 1
net.ipv4.conf.all.accept_source_route = 0
net.ipv6.conf.all.accept_source_route = 0
net.ipv4.conf.default.accept_source_route = 0
net.ipv6.conf.default.accept_source_route = 0
net.ipv4.conf.default.send_redirects = 0
net.ipv4.conf.all.send_redirects = 0
net.ipv4.conf.all.log_martians = 1
net.ipv4.conf.default.log_martians = 1
net.ipv4.conf.all.accept_redirects = 0
net.ipv6.conf.all.accept_redirects = 0
net.ipv4.conf.default.accept_redirects = 0
net.ipv6.conf.default.accept_redirects = 0
{% for item in query('dict', ansible_devices) %}
{% if not item.key.startswith('dm-') %}
{% if item.value.support_discard is defined and item.value.support_discard != "0" %}
[disk_{{ item.key }}]
type=disk
devices={{ item.key }}
elevator=none
{% else %}
[disk_{{ item.key }}]
type=disk
devices={{ item.key }}
elevator=cfq
{% endif %}
{% endif %}
{% endfor %}
It outputs to the following.
[main]
summary=Customized balanced-power usage profile
include=balanced
[bootloader]
cmdline=pcie_aspm=force
[sysctl]
net.ipv4.tcp_rfc1337 = 1
net.ipv4.conf.all.accept_source_route = 0
net.ipv6.conf.all.accept_source_route = 0
net.ipv4.conf.default.accept_source_route = 0
net.ipv6.conf.default.accept_source_route = 0
net.ipv4.conf.default.send_redirects = 0
net.ipv4.conf.all.send_redirects = 0
net.ipv4.conf.all.log_martians = 1
net.ipv4.conf.default.log_martians = 1
net.ipv4.conf.all.accept_redirects = 0
net.ipv6.conf.all.accept_redirects = 0
net.ipv4.conf.default.accept_redirects = 0
net.ipv6.conf.default.accept_redirects = 0
[disk_nvme0n1]
type=disk
devices=nvme0n1
elevator=none
But then complains that tuned.plugins.base: instance disk_nvme0n1: no matching devices available.
2018-05-30 14:02:39,382 WARNING tuned.plugins.plugin_video: radeon_powersave is not supported on 'card0'
2018-05-30 14:02:39,407 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-05-30 14:02:39,500 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-05-30 14:02:39,504 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
2018-05-30 14:03:09,602 INFO tuned.daemon.controller: terminating controller
2018-05-30 14:03:09,602 INFO tuned.daemon.daemon: stopping tuning
2018-05-30 14:03:10,582 INFO tuned.daemon.daemon: terminating Tuned, rolling back all changes
2018-05-30 14:03:10,645 INFO tuned.plugins.plugin_bootloader: removing grub2 tuning previously added by Tuned
2018-05-30 14:03:10,647 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu2'
2018-05-30 14:03:10,647 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu3'
2018-05-30 14:03:10,647 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu0'
2018-05-30 14:03:10,647 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu1'
2018-05-30 14:03:10,752 INFO tuned.daemon.application: dynamic tuning is enabled (can be overridden in plugins)
2018-05-30 14:03:10,756 INFO tuned.daemon.daemon: using sleep interval of 1 second(s)
2018-05-30 14:03:10,756 INFO tuned.daemon.daemon: dynamic tuning is enabled (can be overridden by plugins)
2018-05-30 14:03:10,756 INFO tuned.daemon.daemon: using update interval of 10 second(s) (10 times of the sleep interval)
2018-05-30 14:03:10,757 INFO tuned.profiles.loader: loading profile: custom-balanced
2018-05-30 14:03:10,760 INFO tuned.daemon.controller: starting controller
2018-05-30 14:03:10,760 INFO tuned.daemon.daemon: starting tuning
2018-05-30 14:03:10,769 INFO tuned.plugins.base: instance cpu: assigning devices cpu2, cpu3, cpu0, cpu1
2018-05-30 14:03:10,772 INFO tuned.plugins.plugin_cpu: intel_pstate detected
2018-05-30 14:03:10,772 INFO tuned.plugins.base: instance audio: assigning devices snd_hda_intel
2018-05-30 14:03:10,773 INFO tuned.plugins.base: instance video: assigning devices card0
2018-05-30 14:03:10,773 INFO tuned.plugins.base: instance disk: assigning devices dm-6, dm-7, dm-4, dm-5, dm-2, dm-3, dm-0, dm-1
2018-05-30 14:03:10,773 WARNING tuned.plugins.base: instance scsi_host: no matching devices available
2018-05-30 14:03:10,773 WARNING tuned.plugins.base: instance disk_nvme0n1: no matching devices available
2018-05-30 14:03:10,774 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu2', it is not supported
2018-05-30 14:03:10,774 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu3', it is not supported
2018-05-30 14:03:10,774 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu0', it is not supported
2018-05-30 14:03:10,774 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu1', it is not supported
2018-05-30 14:03:10,777 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu2'
2018-05-30 14:03:10,783 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu3'
2018-05-30 14:03:10,788 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu0'
2018-05-30 14:03:10,794 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu1'
2018-05-30 14:03:10,796 INFO tuned.plugins.plugin_cpu: setting new cpu latency 100
2018-05-30 14:03:10,797 ERROR tuned.utils.commands: Error when reading file '/sys/class/drm/card0/device/power_method': '[Errno 2] No such file or directory: u'/sys/class/drm/card0/device/power_method''
2018-05-30 14:03:10,798 WARNING tuned.plugins.plugin_video: radeon_powersave is not supported on 'card0'
2018-05-30 14:03:10,823 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-05-30 14:03:10,915 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-05-30 14:03:10,919 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
2018-05-30 14:05:51,117 INFO tuned.plugins.plugin_cpu: setting new cpu latency 1000
2018-05-30 14:06:21,156 INFO tuned.plugins.plugin_cpu: setting new cpu latency 100
"/var/log/tuned/tuned.log" [readonly] 560L, 58949C
While that disk is really there.
# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
nvme0n1 259:0 0 238,5G 0 disk
├─nvme0n1p1 259:1 0 100M 0 part /boot/efi
├─nvme0n1p2 259:2 0 1G 0 part /boot
└─nvme0n1p3 259:3 0 237,4G 0 part
The current elevator is set to the following, when I change it to 'mq-deadline', tuned-adm won't set it back to 'none'.
$ cat /sys/block/nvme0n1/queue/scheduler
[none] mq-deadline
According to the documentation, my config should be okay. Could this be a bug or am I doing something wrong here?
# ls -l /dev/nvme0n1*
brw-rw----. 1 root disk 259, 0 30 mei 12:17 /dev/nvme0n1
brw-rw----. 1 root disk 259, 1 30 mei 12:17 /dev/nvme0n1p1
brw-rw----. 1 root disk 259, 2 30 mei 12:17 /dev/nvme0n1p2
brw-rw----. 1 root disk 259, 3 30 mei 12:17 /dev/nvme0n1p3
When I add a '!' as a prefix, the error is gone, but this means the disk is excluded, correct?
What I meant was to try having both the [disk_nvme0n1] and [disk] sections in the config:
[disk_nvme0n1]
type=disk
devices=nvme0n1
elevator=none
[disk]
devices=!nvme0n1
The sections are called "instances" (of the disk plugin in this case) in Tuned terminology. What I think the problem is, is that the profile you're including in your custom profile (powersave or balanced) already defines an instance of the disk plugin called "disk":
[disk]
# Comma separated list of devices, all devices if commented out.
# devices=sda
The problem is that this instance will take over all disk devices in your system, so there will be no devices left for your disk_nvme0n1 instance. The solution would be to override the included instance, so that it doesn't "eat" the nvme0n1 device, hence:
[disk]
devices=!nvme0n1
Below is the log when I comment the include of 'balanced' in my custom config.
2018-05-30 16:25:05,411 INFO tuned.daemon.daemon: stopping tuning
2018-05-30 16:25:05,971 INFO tuned.plugins.plugin_bootloader: removing grub2 tuning previously added by Tuned
2018-05-30 16:25:05,973 INFO tuned.profiles.loader: loading profile: custom-balanced
2018-05-30 16:25:05,974 INFO tuned.daemon.daemon: starting tuning
2018-05-30 16:25:05,979 WARNING tuned.plugins.base: instance disk_nvme0n1: no matching devices available
2018-05-30 16:25:05,979 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-05-30 16:25:06,073 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-05-30 16:25:06,077 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
Below is the log when I still include 'balanced', but include your suggestion as well.
2018-05-30 16:26:52,992 INFO tuned.daemon.daemon: stopping tuning
2018-05-30 16:26:53,276 INFO tuned.plugins.plugin_bootloader: removing grub2 tuning previously added by Tuned
2018-05-30 16:26:53,278 INFO tuned.profiles.loader: loading profile: custom-balanced
2018-05-30 16:26:53,279 INFO tuned.daemon.daemon: starting tuning
2018-05-30 16:26:53,286 INFO tuned.plugins.base: instance cpu: assigning devices cpu2, cpu3, cpu0, cpu1
2018-05-30 16:26:53,289 INFO tuned.plugins.plugin_cpu: intel_pstate detected
2018-05-30 16:26:53,289 INFO tuned.plugins.base: instance audio: assigning devices snd_hda_intel
2018-05-30 16:26:53,290 INFO tuned.plugins.base: instance video: assigning devices card0
2018-05-30 16:26:53,290 INFO tuned.plugins.base: instance disk: assigning devices dm-6, dm-7, dm-4, dm-5, dm-2, dm-3, dm-0, dm-1
2018-05-30 16:26:53,290 WARNING tuned.plugins.base: instance scsi_host: no matching devices available
2018-05-30 16:26:53,290 WARNING tuned.plugins.base: instance disk_nvme0n1: no matching devices available
2018-05-30 16:26:53,291 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu2', it is not supported
2018-05-30 16:26:53,291 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu3', it is not supported
2018-05-30 16:26:53,291 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu0', it is not supported
2018-05-30 16:26:53,291 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu1', it is not supported
2018-05-30 16:26:53,294 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu2'
2018-05-30 16:26:53,300 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu3'
2018-05-30 16:26:53,306 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu0'
2018-05-30 16:26:53,311 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu1'
2018-05-30 16:26:53,314 INFO tuned.plugins.plugin_cpu: setting new cpu latency 100
2018-05-30 16:26:53,315 ERROR tuned.utils.commands: Error when reading file '/sys/class/drm/card0/device/power_method': '[Errno 2] No such file or directory: u'/sys/class/drm/card0/device/power_method''
2018-05-30 16:26:53,315 WARNING tuned.plugins.plugin_video: radeon_powersave is not supported on 'card0'
2018-05-30 16:26:53,341 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-05-30 16:26:53,437 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-05-30 16:26:53,442 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
The warning in the logs persists, even when the 'balanced' profile or disk section in the 'balanced' profile is excluded. The elevator in /sys/block/nvme0n1/queue/scheduler remains unaffected.
Hmm, I didn't expect that. Right now this looks like a bug to me. I'll take a look at this later.
What I missed before is that the device was not even assigned to the "disk" instance from the balanced profile:
2018-05-30 14:03:10,773 INFO tuned.plugins.base: instance disk: assigning devices dm-6, dm-7, dm-4, dm-5, dm-2, dm-3, dm-0, dm-1
I think Tuned doesn't recognize these NVMe drives as supported. Could you provide the output of the following Python script? It gathers some basic info about the device.
from pyudev import Context, Devices
c = Context()
d = Devices.from_path(c, '/sys/block/nvme0n1')
print("Device type: %s" % d.device_type)
print("Removable: %s" % d.attributes.get("removable", None))
if d.parent is None:
print("No subsystem")
else:
print("Subsystem: %s" % d.parent.subsystem)
The output of the script is as follows:
$ python test.py
Device type: disk
Removable: 0
Subsystem: nvme
I also ran the Ansible template on another machine, it also seem to happen with other type of disks.
lsblk output
# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 223,6G 0 disk
├─sda1 8:1 0 100M 0 part /boot/efi
├─sda2 8:2 0 1G 0 part /boot
└─sda3 8:3 0 222,5G 0 part
├─fedora_sovereign-root 253:0 0 112,5G 0 lvm /
├─fedora_sovereign-swap 253:1 0 10G 0 lvm [SWAP]
└─fedora_sovereign-var 253:5 0 100G 0 lvm /var
sdb 8:16 0 1,8T 0 disk
└─sdb1 8:17 0 1,8T 0 part
└─md0 9:0 0 5,5T 0 raid5
├─raid5-home 253:2 0 300G 0 lvm /home
├─raid5-backups 253:3 0 1T 0 lvm /srv/backups
└─raid5-library 253:4 0 4,2T 0 lvm /srv/libray
sdc 8:32 0 1,8T 0 disk
└─sdc1 8:33 0 1,8T 0 part
└─md0 9:0 0 5,5T 0 raid5
├─raid5-home 253:2 0 300G 0 lvm /home
├─raid5-backups 253:3 0 1T 0 lvm /srv/backups
└─raid5-library 253:4 0 4,2T 0 lvm /srv/libray
sdd 8:48 0 1,8T 0 disk
└─sdd1 8:49 0 1,8T 0 part
└─md0 9:0 0 5,5T 0 raid5
├─raid5-home 253:2 0 300G 0 lvm /home
├─raid5-backups 253:3 0 1T 0 lvm /srv/backups
└─raid5-library 253:4 0 4,2T 0 lvm /srv/libray
sde 8:64 0 1,8T 0 disk
└─sde1 8:65 0 1,8T 0 part
└─md0 9:0 0 5,5T 0 raid5
├─raid5-home 253:2 0 300G 0 lvm /home
├─raid5-backups 253:3 0 1T 0 lvm /srv/backups
└─raid5-library 253:4 0 4,2T 0 lvm /srv/libray
tuned profile
[main]
summary=Customized balanced-power usage profile
include=balanced
[bootloader]
cmdline=pcie_aspm=force
[sysctl]
net.ipv4.tcp_rfc1337 = 1
net.ipv4.conf.all.accept_source_route = 0
net.ipv6.conf.all.accept_source_route = 0
net.ipv4.conf.default.accept_source_route = 0
net.ipv6.conf.default.accept_source_route = 0
net.ipv4.conf.default.send_redirects = 0
net.ipv4.conf.all.send_redirects = 0
net.ipv4.conf.all.log_martians = 1
net.ipv4.conf.default.log_martians = 1
net.ipv4.conf.all.accept_redirects = 0
net.ipv6.conf.all.accept_redirects = 0
net.ipv4.conf.default.accept_redirects = 0
net.ipv6.conf.default.accept_redirects = 0
[disk_sdi]
type=disk
devices=sdi
elevator=cfq
[disk_sdd]
type=disk
devices=sdd
elevator=cfq
[disk_sde]
type=disk
devices=sde
elevator=cfq
[disk_sdf]
type=disk
devices=sdf
elevator=cfq
[disk_sdg]
type=disk
devices=sdg
elevator=cfq
[disk_sda]
type=disk
devices=sda
elevator=none
[disk_sdb]
type=disk
devices=sdb
elevator=cfq
[disk_sdc]
type=disk
devices=sdc
elevator=cfq
[disk_sdh]
type=disk
devices=sdh
elevator=cfq
tuned.log
2018-06-04 19:13:49,361 INFO tuned.daemon.daemon: stopping tuning
2018-06-04 19:13:50,239 INFO tuned.plugins.plugin_bootloader: removing grub2 tuning previously added by Tuned
2018-06-04 19:13:50,243 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu6'
2018-06-04 19:13:50,244 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu7'
2018-06-04 19:13:50,244 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu4'
2018-06-04 19:13:50,244 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu5'
2018-06-04 19:13:50,245 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu2'
2018-06-04 19:13:50,245 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu3'
2018-06-04 19:13:50,245 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu0'
2018-06-04 19:13:50,246 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu1'
2018-06-04 19:13:50,247 INFO tuned.profiles.loader: loading profile: custom-balanced
2018-06-04 19:13:50,252 INFO tuned.daemon.daemon: starting tuning
2018-06-04 19:13:50,271 INFO tuned.plugins.base: instance cpu: assigning devices cpu6, cpu7, cpu4, cpu5, cpu2, cpu3, cpu0, cpu1
2018-06-04 19:13:50,275 INFO tuned.plugins.plugin_cpu: intel_pstate detected
2018-06-04 19:13:50,275 WARNING tuned.plugins.base: instance audio: no matching devices available
2018-06-04 19:13:50,275 INFO tuned.plugins.base: instance video: assigning devices card0
2018-06-04 19:13:50,276 INFO tuned.plugins.base: instance disk: assigning devices dm-3, sdd, sde, sda, sdb, sdc, md0, dm-5, dm-2, dm-4, dm-0, dm-1
2018-06-04 19:13:50,276 INFO tuned.plugins.base: instance scsi_host: assigning devices host6, host5, host4, host3, host2, host1, host0
2018-06-04 19:13:50,276 WARNING tuned.plugins.base: instance disk_sdi: no matching devices available
2018-06-04 19:13:50,276 WARNING tuned.plugins.base: instance disk_sdd: no matching devices available
2018-06-04 19:13:50,277 WARNING tuned.plugins.base: instance disk_sde: no matching devices available
2018-06-04 19:13:50,277 WARNING tuned.plugins.base: instance disk_sdf: no matching devices available
2018-06-04 19:13:50,277 WARNING tuned.plugins.base: instance disk_sdg: no matching devices available
2018-06-04 19:13:50,277 WARNING tuned.plugins.base: instance disk_sda: no matching devices available
2018-06-04 19:13:50,278 WARNING tuned.plugins.base: instance disk_sdb: no matching devices available
2018-06-04 19:13:50,278 WARNING tuned.plugins.base: instance disk_sdc: no matching devices available
2018-06-04 19:13:50,278 WARNING tuned.plugins.base: instance disk_sdh: no matching devices available
2018-06-04 19:13:50,279 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu6', it is not supported
2018-06-04 19:13:50,279 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu7', it is not supported
2018-06-04 19:13:50,279 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu4', it is not supported
2018-06-04 19:13:50,279 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu5', it is not supported
2018-06-04 19:13:50,279 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu2', it is not supported
2018-06-04 19:13:50,280 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu3', it is not supported
2018-06-04 19:13:50,280 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu0', it is not supported
2018-06-04 19:13:50,280 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu1', it is not supported
2018-06-04 19:13:50,284 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu6'
2018-06-04 19:13:50,287 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu6, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,289 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu7'
2018-06-04 19:13:50,292 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu7, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,296 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu4'
2018-06-04 19:13:50,299 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu4, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,304 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu5'
2018-06-04 19:13:50,308 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu5, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,312 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu2'
2018-06-04 19:13:50,315 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu2, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,318 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu3'
2018-06-04 19:13:50,321 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu3, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,324 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu0'
2018-06-04 19:13:50,327 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu0, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,330 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu1'
2018-06-04 19:13:50,333 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu1, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-04 19:13:50,333 INFO tuned.plugins.plugin_cpu: setting new cpu latency 1000
2018-06-04 19:13:50,334 ERROR tuned.utils.commands: Error when reading file '/sys/class/drm/card0/device/power_method': '[Errno 2] No such file or directory: u'/sys/class/drm/card0/device/power_method''
2018-06-04 19:13:50,334 WARNING tuned.plugins.plugin_video: radeon_powersave is not supported on 'card0'
2018-06-04 19:13:50,424 INFO tuned.plugins.plugin_scsi_host: ALPM control file ('/sys/class/scsi_host/host6/link_power_management_policy') not found, skipping ALPM setting for 'host6'
2018-06-04 19:13:50,426 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-06-04 19:13:50,545 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-06-04 19:13:50,549 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
Based on the config, /dev/sda should have the 'none' elevator
# cat /sys/block/sda/queue/scheduler
noop deadline [cfq]
The output of the script is as follows:
$ python test.py Device type: disk Removable: 0 Subsystem: nvme
Sorry for the late reply. You should be able to make it work by applying this patch:
diff --git a/tuned/plugins/plugin_disk.py b/tuned/plugins/plugin_disk.py
index 1de1257..6283da0 100644
--- a/tuned/plugins/plugin_disk.py
+++ b/tuned/plugins/plugin_disk.py
@@ -41,7 +41,8 @@ class DiskPlugin(hotplug.Plugin):
return device.device_type == "disk" and \
device.attributes.get("removable", None) == "0" and \
(device.parent is None or \
- device.parent.subsystem in ["scsi", "virtio", "xen"])
+ device.parent.subsystem in \
+ ["scsi", "virtio", "xen", "nvme"])
def _hardware_events_init(self):
self._hardware_inventory.subscribe(self, "block", self._hardware_events_callback)
But before merging this upstream, we'll have to check that the disk plugin can work with NVMe devices.
Based on the config, /dev/sda should have the 'none' elevator
What I said before still holds. You need to override the disk plugin instance from the balanced profile that you're including. From the log i can see that the instance from the balanced profile took over the device:
2018-06-04 19:13:50,276 INFO tuned.plugins.base: instance disk: assigning devices dm-3, sdd, sde, sda, sdb, sdc, md0, dm-5, dm-2, dm-4, dm-0, dm-1
Add e.g. the following to your profile:
[disk]
devices=!sda,!sdb,!sdc,!sdb,!sde
It works!
2018-06-06 12:24:51,486 INFO tuned.daemon.daemon: stopping tuning
2018-06-06 12:24:51,819 INFO tuned.plugins.plugin_bootloader: removing grub2 tuning previously added by Tuned
2018-06-06 12:24:51,823 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu2'
2018-06-06 12:24:51,823 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu3'
2018-06-06 12:24:51,824 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu0'
2018-06-06 12:24:51,825 INFO tuned.plugins.plugin_cpu: setting governor 'powersave' on cpu 'cpu1'
2018-06-06 12:24:51,827 INFO tuned.profiles.loader: loading profile: custom-balanced
2018-06-06 12:24:51,832 INFO tuned.daemon.daemon: starting tuning
2018-06-06 12:24:51,854 INFO tuned.plugins.base: instance cpu: assigning devices cpu2, cpu3, cpu0, cpu1
2018-06-06 12:24:51,864 INFO tuned.plugins.plugin_cpu: intel_pstate detected
2018-06-06 12:24:51,866 INFO tuned.plugins.base: instance audio: assigning devices snd_hda_intel
2018-06-06 12:24:51,867 INFO tuned.plugins.base: instance video: assigning devices card0
2018-06-06 12:24:51,868 WARNING tuned.plugins.base: instance disk: no matching devices available
2018-06-06 12:24:51,869 WARNING tuned.plugins.base: instance scsi_host: no matching devices available
2018-06-06 12:24:51,870 INFO tuned.plugins.base: instance disk_nvme0n1 (disk): assigning devices nvme0n1
2018-06-06 12:24:51,872 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu2', it is not supported
2018-06-06 12:24:51,873 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu3', it is not supported
2018-06-06 12:24:51,873 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu0', it is not supported
2018-06-06 12:24:51,874 INFO tuned.plugins.plugin_cpu: ignoring governor 'conservative' on cpu 'cpu1', it is not supported
2018-06-06 12:24:51,884 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu2'
2018-06-06 12:24:51,894 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu2, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-06 12:24:51,905 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu3'
2018-06-06 12:24:51,916 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu3, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-06 12:24:51,926 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu0'
2018-06-06 12:24:51,934 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu0, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-06 12:24:51,944 INFO tuned.plugins.plugin_cpu: setting energy_perf_bias 'normal' on cpu 'cpu1'
2018-06-06 12:24:51,954 ERROR tuned.utils.commands: Executing x86_energy_perf_policy error: x86_energy_perf_policy: pwrite(cpu1, offset 0x1b0, 0x6) = -1: Operation not permitted
2018-06-06 12:24:51,955 INFO tuned.plugins.plugin_cpu: setting new cpu latency 100
2018-06-06 12:24:51,957 ERROR tuned.utils.commands: Error when reading file '/sys/class/drm/card0/device/power_method': '[Errno 2] No such file or directory: u'/sys/class/drm/card0/device/power_method''
2018-06-06 12:24:51,958 WARNING tuned.plugins.plugin_video: radeon_powersave is not supported on 'card0'
2018-06-06 12:24:51,958 INFO tuned.plugins.plugin_bootloader: installing additional boot command line parameters to grub2
2018-06-06 12:24:52,136 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2018-06-06 12:24:52,159 INFO tuned.daemon.daemon: static tuning from profile 'custom-balanced' applied
And yes, I forgot about excluding those disks again. Just for those who might stumble upon this thread, the Ansible template looks like this to include the exclusion of the devices in the disk section.
[main]
summary={{ item.summary }}
include={{ item.include }}
[bootloader]
cmdline=pcie_aspm=force
[sysctl]
net.ipv4.tcp_rfc1337 = 1
net.ipv4.conf.all.accept_source_route = 0
net.ipv6.conf.all.accept_source_route = 0
net.ipv4.conf.default.accept_source_route = 0
net.ipv6.conf.default.accept_source_route = 0
net.ipv4.conf.default.send_redirects = 0
net.ipv4.conf.all.send_redirects = 0
net.ipv4.conf.all.log_martians = 1
net.ipv4.conf.default.log_martians = 1
net.ipv4.conf.all.accept_redirects = 0
net.ipv6.conf.all.accept_redirects = 0
net.ipv4.conf.default.accept_redirects = 0
net.ipv6.conf.default.accept_redirects = 0
[disk]
devices=!{{ ansible_devices.keys() | join(',!') }}
{% for item in query('dict', ansible_devices) if not item.key.startswith('dm-') and not item.key.startswith('md') %}
{% if item.value.support_discard is defined and item.value.support_discard != "0" %}
[disk_{{ item.key }}]
type=disk
devices={{ item.key }}
elevator=none
{% else %}
[disk_{{ item.key }}]
type=disk
devices={{ item.key }}
elevator=cfq
{% endif %}
{% endfor %}
Hi,
I'm getting the same problem, using tuned 2.10 (or 2.13 from sources) on Debian 9.
It's using a virtual machine with virtio.
The test.py gets (for vda):
root@EVO:~/tuned-2.13.0# python test.py
Device type: disk
Removable: 0
Subsystem: virtio
But tuned still says:
2020-02-26 10:23:18,914 WARNING tuned.plugins.base: instance disk: no matching devices available 2020-02-26 10:23:18,914 WARNING tuned.plugins.base: instance disk-all: no matching devices available
I'm using virtual-guest.
2020-02-26 11:46:22,939 INFO tuned.profiles.loader: loading profile: virtual-guest
2020-02-26 11:46:22,940 INFO tuned.daemon.daemon: starting tuning
2020-02-26 11:46:22,942 INFO tuned.plugins.base: instance cpu: assigning devices cpu1, cpu0
2020-02-26 11:46:22,944 WARNING tuned.plugins.plugin_cpu: unable to run x86_energy_perf_policy tool, ignoring CPU energy performance bias, is the tool installed?
2020-02-26 11:46:22,944 WARNING tuned.plugins.base: instance disk: no matching devices available
2020-02-26 11:46:22,945 INFO tuned.plugins.plugin_cpu: setting new cpu latency 1000
2020-02-26 11:46:22,968 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl
2020-02-26 11:46:22,970 INFO tuned.daemon.daemon: static tuning from profile 'virtual-guest' applied
The [disk] from "trhoughput-performance" which is including is creating the "2020-02-26 11:46:22,944 WARNING tuned.plugins.base: instance disk: no matching devices available".
I'm using virtual-guest.
Unmodified?
That's odd. I don't immediatelly see what could be wrong.
If you're up to some debugging, I suggest you use pdb and take a look at what's happenning here:
https://github.com/redhat-performance/tuned/blob/9bb73e8f407ae3ce17031e020ab112ac8c8cef45/tuned/plugins/plugin_disk.py#L31-L33
@ouafnico Can you please provide the output of this script?
import pyudev
ctx = pyudev.Context()
for i in ctx.list_devices(subsystem="block"):
print('%s %s' % (i, i.sys_name))
Sure :)
# python debug.py
Device(u'/sys/devices/pci0000:00/0000:00:02.3/0000:04:00.0/virtio2/block/vda') vda
Device(u'/sys/devices/pci0000:00/0000:00:02.3/0000:04:00.0/virtio2/block/vda/vda1') vda1
Are you running this with the same python version as Tuned uses on Debian?
It's tested with the default python 2.7.13 from debian stretch. If I run it with the same as tuned, 3.5.3, I got this:
# python3 debug.py
Device('/sys/devices/pci0000:00/0000:00:02.3/0000:04:00.0/virtio2/block/vda') vda
Device('/sys/devices/pci0000:00/0000:00:02.3/0000:04:00.0/virtio2/block/vda/vda1') vda1
OK, that's pretty much the same output. What about the output of the previous script? Is there a difference between running it with python 2 and 3?
If not, I'm out of ideas.
Hum :
# python test.py
Device type: disk
Removable: 0
Subsystem: virtio
# python3 test.py
Device type: disk
Removable: b'0'
Subsystem: virtio
I'm out of ideas too...
I remembered that we've fixed this kind of problem in the past: https://github.com/redhat-performance/tuned/commit/2cc3d747986837d7e
That commit is not contained in tuned 2.10, so that could explain why it doesn't work for you with the Debian's Tuned version.
It doesn't explain though why it doesn't work for you with 2.13. But perhaps, due to some mistake, you're not really running the 2.13 version, but the Debian's version?
Can you try applying that patch to the Debian package? The file you need to patch could be somewhere like /usr/lib/python*/site-packages/tuned/plugins/plugin_disk.py.
I have this commit on 2.13. I'm running the good version on debian, but same problem.
I didn't have the original 2.10 package on my debian (only using the sources for 2.13).