Hello,
To run x86 software I found a well priced thin client (45 € used) with a Celeron N3160 processor. It runs the UEFI/SquashFS firmware (generic-squashfs-combined-efi.img.gz).
However, I was having stability issues and the device occasionally restarted itself. What I tried:
-
opkg install "intel-microcode", but still the device restarted - Found a test case that caused the device to freeze/hang-up/restart:
while true; do stress --cpu 8 --io 4 --vm 2 --vm-bytes 1G --timeout 120s --hdd 2; sleep 120; done - disabled all C-states higher C1 in the BIOS, but it still restarts occasionally
- Modified the grub commandline to disable all Intel P-states:
sed -i "s/console=tty0 console=ttyS0,115200n8 noinitrd$/console=tty0 console=ttyS0,115200n8 noinitrd intel_pstate=disable/" /boot/grub/grub.cfgbut it still restarts - Added the following to the
/etc/rc.local:
#Stability improvement, loosing powersaving options
sed -i "s/console=tty0 console=ttyS0,115200n8 noinitrd$/console=tty0 console=ttyS0,115200n8 noinitrd intel_pstate=disable/" /boot/grub/grub.cfg
for CORE in 0 1 2 3; do
echo "echo performance > /sys/devices/system/cpu/cpu$CORE/cpufreq/scaling_governor"
echo performance > /sys/devices/system/cpu/cpu$CORE/cpufreq/scaling_governor
done
MAXFREQ="$(cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq)"
MAXFREQ="1600000"
MINFREQ="1600000"
for CORE in 0 1 2 3; do
echo "echo $MAXFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_max_freq"
echo $MAXFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_max_freq
echo "echo $MINFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_min_freq"
echo $MINFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_min_freq
done
#get all c-states: grep . /sys/devices/system/cpu/cpu*/cpuidle/state*/name
for CORE in 0 1 2 3; do
for STATE in 2 3 4 5; do
echo "echo 1 > /sys/devices/system/cpu/cpu$CORE/cpuidle/state$STATE/disable"
echo 1 > /sys/devices/system/cpu/cpu$CORE/cpuidle/state$STATE/disable
done
done
Now it seems to keep running, the test humms along quite reliably and without restarts/freezes.
But: I sacrificed most power saving options in exchange for stability. Perhaps you have suggestions what I might do instead.
Here a few graphs how it now performs with the test script:
This picture shows the CPU load of all cores over the past hour. Quite obvious is the stop and go of the test script:
The frequency remains at 1.6 GHz all the time. Before it switched frequencies and saved power, but it chrashed. With fixed settings it does not reboot anymore by itself:
The CPU temperature remains acceptable:
Edit 02.03.2022:
Experimenting with settings does not cause reboots with scaling_governor "ondemand" and frequencies not lower than 1.2 GHz. This is now the most recent /etc/rc.local:
#Stability improvement
sed -i "s/console=tty0 console=ttyS0,115200n8 noinitrd$/console=tty0 console=ttyS0,115200n8 noinitrd intel_pstate=disable/" /boot/grub/grub.cfg
for CORE in 0 1 2 3; do
echo "echo ondemand > /sys/devices/system/cpu/cpu$CORE/cpufreq/scaling_governor"
echo ondemand > /sys/devices/system/cpu/cpu$CORE/cpufreq/scaling_governor
done
MAXFREQ="$(cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq)"
#MAXFREQ="1600000"
#MINFREQ="1600000"
MINFREQ="1200000"
#MINFREQ="800000"
#MINFREQ="480000"
for CORE in 0 1 2 3; do
echo "echo $MAXFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_max_freq"
echo $MAXFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_max_freq
echo "echo $MINFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_min_freq"
echo $MINFREQ > /sys/devices/system/cpu/cpufreq/policy$CORE/scaling_min_freq
done
# get all c-states:
# bash -c "grep . /sys/devices/system/cpu/cpu*/cpuidle/state*/{name,disable} | sort"
# 0 -> POLL,
# 1 -> C1,
# 2 -> C6N,
# 3 -> C6S,
# 4 -> C7,
# 5 -> C7S
for CORE in 0 1 2 3; do
for STATE in 2 3 4 5; do
echo "echo 1 > /sys/devices/system/cpu/cpu$CORE/cpuidle/state$STATE/disable"
echo 1 > /sys/devices/system/cpu/cpu$CORE/cpuidle/state$STATE/disable
done
done
#opkg update
#opkg install fdisk block-mount lsblk htop collectd-mod-ping collectd-mod-wireless luci-app-statistics pingcheck collectd-mod-uptime \
# kmod-usb-storage e2fsprogs fdisk resize2fs \
# htop debootstrap debian-archive-keyring gpgv2 block-mount \
# qemu-img qemu-x86_64-softmmu kmod-kvm-x86 qemu-bridge-helper kmod-tun \
# openssh-sftp-server bash screen
#run Paperless-NG or the successor Paperless-NGX, keep it in a screen session check how it is doing
screen -d -S paperless -m bash /etc/config/Paperless-NG/99\ run\ paperless.ng.sh
exit 0
