MicroShift and KubeVirt on Raspberry Pi 4 with Manjaro 22.04
Introduction
MicroShift is a research project that is exploring how OpenShift OKD Kubernetes distribution can be optimized for small form factor devices and edge computing. In Part 1 we looked at multiple ways to run MicroShift on a MacBook Pro. In Part 4, we ran MicroShift on the Raspberry Pi 4 with the Raspberry Pi OS (64 bit) and further in Part 9, we looked at Virtualization with MicroShift on Raspberry Pi 4 with Raspberry Pi OS (64 bit). In Part 5, we saw multiple options to build and run MicroShift on the Raspberry Pi 4 with the CentOS 8 Stream. In Part 6, we deployed MicroShift on the Raspberry Pi 4 with Ubuntu 20.04. In Part 8, we looked at the All-In-One install of MicroShift on balenaOS. In Part 10, Part 11, and Part 12, we deployed MicroShift and KubeVirt on Fedora IoT, Fedora Server and Fedora CoreOS respectively, Part 13 with Ubuntu 22.04, Part 14 on Rocky Linux, Part 15 on openSUSE, Part 16 on Oracle Linux and Part 17 on AlmaLinux. In this Part 18, we will work with MicroShift on Manjaro (based on ArchLinux). We will run an object detection sample and send messages to Node Red installed on MicroShift. Further, we will setup KubeVirt and the OKD Web Console and run Virtual Machine Instances in MicroShift. We will also run sample notebooks for object detection and license plate detection.
Manjaro inherits a lot of the features of Arch Linux and implements many unique features. It subscribes to the rolling release development model; updates are continuous and it’s never necessary to download a new release of Manjaro. You keep your system up to date through the Pacman package manager. Manjaro is only available for 64 bit CPU architectures. Manjaro has several editions, including “official” and “community”, which both contain several different supported desktop environments. Support for AppArmor, SELinux and Tomoyo was removed in the Arch kernel, and this also applies to the Manjaro kernels. Manjaro also supports ARM and has a download tailored to Raspberry Pi systems. We use the Minimal image without the desktop environment.
Setting up the Raspberry Pi 4 with Manjaro
Run the following steps to download the Manjaro image and setup the Raspberry Pi 4.
1. Download the Manjaro Raspberry Pi 4 Minimal from https://manjaro.org/download/
2. Write to Microsdxc card using balenaEtcher or the Raspberry Pi Imager
3. Have a Keyboard and Monitor connected to the Raspberry Pi 4
4. Insert Microsdxc into Raspberry Pi 4 and poweron
5. Set the user rpi, hostname rpi and select the other inputs as required using the Keyboard and Monitor attached to the Raspberry Pi. It will automatically resize the partition and reboot.
6. Find the ethernet dhcp ipaddress of your Raspberry Pi 4 by running the nmap on your Macbook with your subnet
$ sudo nmap -sn 192.168.1.0/24
Nmap scan report for 192.168.1.228
Host is up (0.080s latency).
MAC Address: E4:5F:01:2E:D8:95 (Raspberry Pi Trading)
7. Login using rpi and password (you set above) to ipaddress of your Raspberry Pi.
$ ssh rpi@192.168.1.228
rpi@192.168.1.228's password:
Welcome to Manjaro-ARM
~~Website: https://manjaro.org
~~Forum: https://forum.manjaro.org/c/arm
~~Matrix: #manjaro-arm-public:matrix.org
[rpi@rpi ~]$ sudo su -
We trust you have received the usual lecture from the local System
Administrator. It usually boils down to these three things:
#1) Respect the privacy of others.
#2) Think before you type.
#3) With great power comes great responsibility.
[sudo] password for rpi:
[root@rpi ~]#
8. Check the partition is using the full size
[root@rpi ~]# df -h
Filesystem Size Used Avail Use% Mounted on
dev 3.7G 0 3.7G 0% /dev
run 3.9G 13M 3.9G 1% /run
/dev/mmcblk0p2 58G 1.5G 54G 3% /
tmpfs 3.9G 0 3.9G 0% /dev/shm
tmpfs 3.9G 0 3.9G 0% /tmp
/dev/mmcblk0p1 214M 56M 159M 26% /boot
tmpfs 782M 0 782M 0% /run/user/1000
9. Update and add the ipv4 address to /etc/hosts
yes Y | sudo pacman -Syyu
hostnamectl set-hostname rpi.example.com
echo "$ipaddress rpi rpi.example.com" >> /etc/hosts
10. Optionally, enable wifi using nmcli
yes Y | pacman -S networkmanager
systemctl enable --now NetworkManager
sleep 10
nmcli device wifi list # Note your ssid
nmcli device wifi connect $ssid --ask
11. Check the release
cat /etc/os-release
[root@rpi ~]# cat /etc/os-release
NAME="Manjaro-ARM"
ID="manjaro-arm"
ID_LIKE="manjaro arch"
PRETTY_NAME="Manjaro ARM"
ANSI_COLOR="1;32"
HOME_URL="https://www.manjaro.org/"
SUPPORT_URL="https://forum.manjaro.org/c/arm/"
LOGO=manjarolinux
12. Update the kernel parameters
We want the memory to be enabled, it is not enabled by default
[root@rpi ~]# mount | grep cgroup # Check that memory and cpuset are present
cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,nsdelegate,memory_recursiveprot)
[root@rpi ~]# cat /proc/cgroups | column -t # Check that memory and cpuset are present
#subsys_name hierarchy num_cgroups enabled
cpuset 0 32 1
cpu 0 32 1
cpuacct 0 32 1
blkio 0 32 1
memory 0 32 0
devices 0 32 1
freezer 0 32 1
net_cls 0 32 1
perf_event 0 32 1
net_prio 0 32 1
pids 0 32 1
rdma 0 32 1
Edit the cmdline.txt and reboot
yes Y | pacman -S vim vi
vim /boot/cmdline.txt
Concatenate the following onto the end of the existing line (do not add a new line) in /boot/cmdline.txt
cgroup_enable=cpuset cgroup_memory=1 cgroup_enable=memory
A control group (cgroup) is a Linux kernel feature that limits, accounts for, and isolates the resource usage (CPU, memory, disk I/O, network, and so on) of a collection of processes. Cgroups are a key component of containers because there are often multiple processes running in a container that you need to control together. In Microshift, cgroups are used to implement resource requests and limits and corresponding QoS classes at the pod level.
reboot
Verify
ssh rpi@$ipaddress
sudo su -
cat /proc/cmdline
mount | grep cgroup # Check that memory and cpuset are present
cat /proc/cgroups | column -t # Check that memory and cpuset are present
Output:
[root@rpi ~]# cat /proc/cmdline
coherent_pool=1M 8250.nr_uarts=0 snd_bcm2835.enable_compat_alsa=0 snd_bcm2835.enable_hdmi=1 video=HDMI-A-1:1920x1200M@60 smsc95xx.macaddr=E4:5F:01:2E:D8:95 vc_mem.mem_base=0x3ec00000 vc_mem.mem_size=0x40000000 root=PARTUUID=1e7027b0-02 rw rootwait console=ttyS0,115200 console=tty3 selinux=0 quiet splash plymouth.ignore-serial-consoles smsc95xx.turbo_mode=N dwc_otg.lpm_enable=0 kgdboc=ttyS0,115200 usbhid.mousepoll=8 audit=0 cgroup_enable=cpuset cgroup_memory=1 cgroup_enable=memory
[root@rpi ~]# mount | grep cgroup # Check that memory and cpuset are present
cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,nsdelegate,memory_recursiveprot)
[root@rpi ~]# cat /proc/cgroups | column -t # Check that memory and cpuset are present
#subsys_name hierarchy num_cgroups enabled
cpuset 0 60 1
cpu 0 60 1
cpuacct 0 60 1
blkio 0 60 1
memory 0 60 1
devices 0 60 1
freezer 0 60 1
net_cls 0 60 1
perf_event 0 60 1
net_prio 0 60 1
pids 0 60 1
rdma 0 60 1
Install sense_hat and RTIMULib on ArchLinux
The Sense HAT is an add-on board for the Raspberry Pi. The Sense HAT has an 8 × 8 RGB LED matrix, a five – button joystick and includes the following sensors: Inertial Measurement Unit (Accelerometer, Gyroscope, Magnetometer), Temperature, Barometric pressure, Humidity. If you have the Sense HAT attached, install the libraries.
Install sensehat
yes Y | pacman -S i2c-tools make cmake gcc python-pip
pip3 install Cython Pillow numpy sense_hat
Check the Sense Hat with i2cdetect
modprobe i2c-dev
modprobe i2c-bcm2708
echo "i2c-dev" > /etc/modules-load.d/i2c-dev.conf
echo "i2c-bcm2708" > /etc/modules-load.d/i2c-bcm2708.conf
i2cdetect -y 1
Output
[root@rpi ~]# modprobe i2c-dev
[root@rpi ~]# modprobe i2c-bcm2708
[root@rpi ~]# i2cdetect -y 1
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- 1c -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- UU -- -- -- -- -- -- -- -- --
50: -- -- -- -- -- -- -- -- -- -- -- -- 5c -- -- 5f
60: -- -- -- -- -- -- -- -- -- -- 6a -- -- -- -- --
70: -- -- -- -- -- -- -- --
Install RTIMULib
yes Y | pacman -S git
cd ~
git clone https://github.com/RPi-Distro/RTIMULib.git
cd RTIMULib/
cd Linux/python
python3 setup.py build
python3 setup.py install
cd ../..
cd RTIMULib
mkdir build
cd build
cmake ..
make -j4
make install
ldconfig
# Optional test the sensors
cd /root/RTIMULib/Linux/RTIMULibDrive11
make -j4
make install
RTIMULibDrive11 # Ctrl-C to break
cd /root/RTIMULib/Linux/RTIMULibDrive10
make -j4
make install
RTIMULibDrive10 # Ctrl-C to break
# Optional
yes Y | pacman -S qt5-base
cd /root/RTIMULib/Linux/RTIMULibDemoGL
qmake-qt5
make -j4
make install
Test the SenseHat samples for the Sense Hat's LED matrix and sensors.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/sensehat-fedora-iot
# Enable random LEDs
python3 sparkles.py # Ctrl-C to interrupt
# Show multiple screens to test LEDs
python3 rainbow.py # Ctrl-C to interrupt
# First time you run the temperature.py, you may see “Temperature: 0 C”. Just run it again.
python3 temperature.py
# Show the Temperature, Pressure and Humidity
python3 testsensehat.py # Ctrl-C to interrupt
# Show two digits for multiple numbers
sed -i "s/32,32,32/255,255,255/" digits.py
python3 digits.py
# When a magnet gets close to SenseHAT, the LEDs will all turn red for 1/5 of a second
python3 magnetometer.py
# Find Magnetic North
python3 compass.py
Install MicroShift on the Raspberry Pi 4 Manjaro host
Install the dependencies and copy the latest microshift prebuilt binary. You will also need to set the crio.conf.
yes Y | pacman -S firewalld cri-o crictl
# Check the registries /etc/crio/crio.conf
#echo 'registries=["docker.io"]' >> /etc/crio/crio.conf
echo 'unqualified-search-registries=["docker.io"]' >> /etc/containers/registries.conf
ARCH=arm64
VERSION=$(curl -sL https://api.github.com/repos/redhat-et/microshift/releases | grep tag_name | head -n 1 | cut -d '"' -f 4)
curl -LO https://github.com/redhat-et/microshift/releases/download/$VERSION/microshift-linux-$ARCH
curl -LO https://github.com/redhat-et/microshift/releases/download/$VERSION/release.sha256
BIN_SHA="$(sha256sum microshift-linux-$ARCH | awk '{print $1}')"
KNOWN_SHA="$(grep "microshift-linux-$ARCH" release.sha256 | awk '{print $1}')"
if [[ "$BIN_SHA" != "$KNOWN_SHA" ]]; then
echo "SHA256 checksum failed" && exit 1
fi
sudo chmod +x microshift-linux-$ARCH
sudo mv microshift-linux-$ARCH /usr/bin/microshift
yes Y | pacman -S --needed wget
wget https://raw.githubusercontent.com/redhat-et/microshift/main/packaging/systemd/microshift.service -O /usr/lib/systemd/system/microshift.service
systemctl daemon-reload
Install KVM on the host and validate the Host Virtualization Setup.
yes Y | pacman -S jack2 virt-manager virt-viewer virt-install libvirt qemu ebtables dnsmasq bridge-utils
If the above command gives error, then we need to download some packages from the Arch x86_64 repo and install them using pacman as mentioned at https://archlinuxarm.org/forum/viewtopic.php?t=16037&p=69506
mkdir packages;cd packages
wget https://archlinux.org/packages/extra/any/edk2-ovmf/download -O edk2-ovmf-202202-2-any.pkg.tar.zst
wget https://archlinux.org/packages/extra/any/seabios/download -O seabios-1.16.0-1-any.pkg.tar.zst
wget http://mirror.archlinuxarm.org/aarch64/extra/qemu-system-x86-7.0.0-10-aarch64.pkg.tar.xz
wget https://archlinux.org/packages/extra/any/edk2-armvirt/download -O edk2-armvirt-202202-2-any.pkg.tar.zst
pacman -U --overwrite \* --noconfirm *
pacman -S jack2 virt-manager virt-viewer virt-install libvirt qemu ebtables dnsmasq bridge-utils qemu-system-aarch64
This will install the qemu-base and remove the iptables as shown below:
[root@rpi packages]# pacman -S jack2 virt-manager virt-viewer virt-install libvirt qemu ebtables dnsmasq bridge-utils qemu-system-aarch64
:: There are 3 providers available for qemu:
:: Repository extra
1) qemu-base 2) qemu-desktop 3) qemu-full
Enter a number (default=1): 1
resolving dependencies...
looking for conflicting packages...
:: iptables-nft and iptables are in conflict. Remove iptables? [y/N] y
warning: dependency cycle detected:
warning: libglvnd will be installed before its mesa dependency
The virt-host-validate command validates that the host is configured in a suitable way to run libvirt hypervisor driver qemu.
# Works with nftables on Fedora Server and Fedora IoT, Oracle Linux, AlmaLinux, ArchLinux
# vim /etc/firewalld/firewalld.conf
systemctl enable --now libvirtd
virt-host-validate qemu
Output:
[root@rpi sensehat-fedora-iot]# systemctl enable --now libvirtd
Created symlink /etc/systemd/system/multi-user.target.wants/libvirtd.service → /usr/lib/systemd/system/libvirtd.service.
Created symlink /etc/systemd/system/sockets.target.wants/virtlockd.socket → /usr/lib/systemd/system/virtlockd.socket.
Created symlink /etc/systemd/system/sockets.target.wants/virtlogd.socket → /usr/lib/systemd/system/virtlogd.socket.
Created symlink /etc/systemd/system/sockets.target.wants/libvirtd.socket → /usr/lib/systemd/system/libvirtd.socket.
Created symlink /etc/systemd/system/sockets.target.wants/libvirtd-ro.socket → /usr/lib/systemd/system/libvirtd-ro.socket.
[root@rpi sensehat-fedora-iot]# virt-host-validate qemu
QEMU: Checking if device /dev/kvm exists : PASS
QEMU: Checking if device /dev/kvm is accessible : PASS
QEMU: Checking if device /dev/vhost-net exists : PASS
QEMU: Checking if device /dev/net/tun exists : PASS
QEMU: Checking for cgroup 'cpu' controller support : PASS
QEMU: Checking for cgroup 'cpuacct' controller support : PASS
QEMU: Checking for cgroup 'cpuset' controller support : PASS
QEMU: Checking for cgroup 'memory' controller support : PASS
QEMU: Checking for cgroup 'devices' controller support : PASS
QEMU: Checking for cgroup 'blkio' controller support : PASS
QEMU: Checking for device assignment IOMMU support : WARN (Unknown if this platform has IOMMU support)
QEMU: Checking for secure guest support : WARN (Unknown if this platform has Secure Guest support)
Check that cni plugins are present
ls /opt/cni/bin/ # cni plugins
ls /usr/libexec/cni # No such file or directory
Output:
[root@rpi microshift]# ls /opt/cni/bin/ # cni plugins
bandwidth bridge dhcp firewall host-device host-local ipvlan loopback macvlan portmap ptp sbr static tuning vlan vrf
[root@rpi microshift]# ls /usr/libexec/cni # empty
ls: cannot access '/usr/libexec/cni': No such file or directory
We will have systemd start and manage MicroShift. Refer to the microshift service for the three approaches.
systemctl enable --now crio microshift
You may read about selecting zones for your interfaces.
systemctl enable firewalld --now
firewall-cmd --zone=trusted --add-source=10.42.0.0/16 --permanent
firewall-cmd --zone=trusted --add-source=10.85.0.0/16 --permanent
firewall-cmd --zone=public --add-port=80/tcp --permanent
firewall-cmd --zone=public --add-port=443/tcp --permanent
firewall-cmd --zone=public --add-port=5353/udp --permanent
firewall-cmd --reload
Additional ports may need to be opened. For external access to run kubectl or oc commands against MicroShift, add the 6443 port:
firewall-cmd --zone=public --permanent --add-port=6443/tcp
For access to services through NodePort, add the port range 30000-32767:
firewall-cmd --zone=public --permanent --add-port=30000-32767/tcp
firewall-cmd --reload
firewall-cmd --list-all --zone=public
firewall-cmd --get-default-zone
#firewall-cmd --set-default-zone=public
#firewall-cmd --get-active-zones
firewall-cmd --list-all
Check the microshift and crio logs
journalctl -u microshift -f
journalctl -u crio -f
The microshift service references the microshift binary in the /usr/bin directory
[root@rpi ~]# cat /usr/lib/systemd/system/microshift.service
[Unit]
Description=MicroShift
Wants=network-online.target crio.service
After=network-online.target crio.service
[Service]
WorkingDirectory=/usr/bin/
ExecStart=microshift run
Restart=always
User=root
[Install]
WantedBy=multi-user.target
Install the kubectl and the openshift oc client
ARCH=arm64
cd /tmp
dnf -y install tar
export OCP_VERSION=4.9.11 && \
curl -o oc.tar.gz https://mirror2.openshift.com/pub/openshift-v4/$ARCH/clients/ocp/$OCP_VERSION/openshift-client-linux-$OCP_VERSION.tar.gz && \
tar -xzvf oc.tar.gz && \
rm -f oc.tar.gz && \
install -t /usr/local/bin {kubectl,oc} && \
rm -f {README.md,kubectl,oc}
It will take around 3 minutes for all pods to start. Check the status of node and pods using kubectl or oc client.
export KUBECONFIG=/var/lib/microshift/resources/kubeadmin/kubeconfig
#watch "kubectl get nodes;kubectl get pods -A;crictl pods;crictl images"
watch "oc get nodes;oc get pods -A;crictl pods;crictl images"
Output of top after MicroShift is started, see the memory usage
top - 04:41:57 up 13 min, 1 user, load average: 1.37, 2.20, 1.34
Tasks: 159 total, 1 running, 158 sleeping, 0 stopped, 0 zombie
%Cpu(s): 6.3 us, 2.0 sy, 0.0 ni, 91.4 id, 0.1 wa, 0.0 hi, 0.2 si, 0.0 st
MiB Mem : 7810.5 total, 5626.7 free, 955.5 used, 1228.3 buff/cache
MiB Swap: 11715.7 total, 11715.7 free, 0.0 used. 6743.2 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
731 root 20 0 13.2g 767016 110208 S 30.3 9.6 5:12.09 microshift
688 root 20 0 1940960 70836 40208 S 1.0 0.9 1:23.65 crio
2069 root 20 0 747776 41956 31776 S 1.0 0.5 0:01.95 coredns
3193 root 20 0 16864 5692 4576 R 1.0 0.1 0:00.66 top
1759 root 20 0 755076 57380 37904 S 0.7 0.7 0:29.37 service-ca-oper
77 root 0 -20 0 0 0 I 0.3 0.0 0:03.08 kworker/u9:0-brcmf_wq/mmc1:0001:1
94 root rt 0 0 0 0 S 0.3 0.0 0:01.16 sugov:0
187 root 20 0 49544 17480 16388 S 0.3 0.2 0:02.11 systemd-journal
1043 root 20 0 0 0 0 D 0.3 0.0 0:00.79 kworker/0:4+events
1421 root 20 0 748840 36080 26300 S 0.3 0.5 0:01.20 flanneld
Install podman - We will use podman for containerized deployment of MicroShift and building images for the samples.
yes Y | pacman -S podman buildah skopeo
Samples to run on MicroShift
1. InfluxDB/Telegraf/Grafana
The source code is available for this influxdb sample in github.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/influxdb
If you want to run all the steps in a single command, get the nodename.
oc get nodes
Output:
[root@rpi influxdb]# oc get nodes
NAME STATUS ROLES AGE VERSION
rpi.example.com Ready
Replace the annotation kubevirt.io/provisionOnNode with the above nodename and execute the runall-balena-dynamic.sh. Note that the node name is different when running MicroShift with the all-in-one containerized approach. So, you will use the microshift.example.com instead of the rpi.example.com.
sed -i "s|coreos|rpi.example.com|" influxdb-data-dynamic.yaml
sed -i "s|coreos|rpi.example.com|" grafana/grafana-data-dynamic.yaml
./runall-balena-dynamic.sh
We create and push the “measure:latest” image using the Dockerfile in the sensor directory. The script will create a new project influxdb for this sample, install InfluxDB, install the pod for SenseHat measurements, install Telegraf and check the measurements for the telegraf database in InfluxDB. Finally, it will install Grafana.
This script will allocate dynamic persistent volumes using influxdb-data-dynamic.yaml and grafana-data-dynamic.yaml. The annotation provisionOnNode and the storageClassName are required for dynamic PV.
annotations:
kubevirt.io/provisionOnNode: rpi.example.com
spec:
storageClassName: kubevirt-hostpath-provisioner
Persistent Volumes and Claims Output:
[root@rpi influxdb]# oc get pv,pvc
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
persistentvolume/pvc-d120a1c7-ab39-4101-bcf4-66e2930aed52 57Gi RWO Delete Bound influxdb/grafana-data kubevirt-hostpath-provisioner 5m5s
persistentvolume/pvc-e4c56850-0739-410d-840e-c45c602da375 57Gi RWO Delete Bound influxdb/influxdb-data kubevirt-hostpath-provisioner 8m9s
NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
persistentvolumeclaim/grafana-data Bound pvc-d120a1c7-ab39-4101-bcf4-66e2930aed52 57Gi RWO kubevirt-hostpath-provisioner 5m6s
persistentvolumeclaim/influxdb-data Bound pvc-e4c56850-0739-410d-840e-c45c602da375 57Gi RWO kubevirt-hostpath-provisioner 8m9s
Add the "<RaspberryPiIPAddress> grafana-service-influxdb.cluster.local" to /etc/hosts on your laptop and login to http://grafana-service-influxdb.cluster.local/login using admin/admin. You will need to change the password on first login. Go to the Dashboards list (left menu > Dashboards > Manage). Open the Analysis Server dashboard to display monitoring information for MicroShift. Open the Balena Sense dashboard to show the temperature, pressure, and humidity from SenseHat.
Finally, after you are done working with this sample, you can run the deleteall-balena-dynamic.sh
./deleteall-balena-dynamic.sh
Deleting the persistent volume claims automatically deletes the persistent volumes.
2. Node Red live data dashboard with SenseHat sensor charts
We will install Node Red on the ARM device as a deployment within MicroShift, add the dashboard and view the gauges for temperature/pressure/humidity data from SenseHat on the dashboard.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/nodered
Build and push the arm64v8 image “karve/nodered:arm64”
cd docker-custom/
# Replace docker with podman in docker-debian.sh and run it
./docker-debian.sh
podman push karve/nodered:arm64
cd ..
Deploy Node Red with persistent volume for /data within the node red container
mkdir /var/hpvolumes/nodered
restorecon -R -v "/var/hpvolumes/*"
rm -rf /var/hpvolumes/nodered/*;cp -r nodered-volume/* /var/hpvolumes/nodered/.
oc new-project nodered
oc apply -f noderedpv.yaml -f noderedpvc.yaml -f nodered3.yaml -f noderedroute.yaml
oc get routes
oc -n nodered wait deployment nodered-deployment --for condition=Available --timeout=300s
oc logs deployment/nodered-deployment -f
Add the ipaddress of the Raspberry Pi 4 device for nodered-svc-nodered.cluster.local to /etc/hosts on your Laptop and browse to http://nodered-svc-nodered.cluster.local/
The following modules required for the dashboard have been preinstalled node-red-dashboard, node-red-node-smooth, node-red-node-pi-sense-hat. These can be seen under “Manage Palette - Install”. The Flow 1 or Flow 2 have already been imported from the nodered sample. This import to the Node Red can be done manually under “Import Nodes” and then click “Deploy”.
Double click the Sense HAT input node and make sure that all the events are checked. Select the Dashboard. Click on the outward arrow in the tabs to view the sensor charts. You will see the Home by Default. You can see the state of the Joystick Up, Down, Left, Right or Pressed. Click on the Hamburger Menu (3 lines) and select PiSenseHAT. If you selected the Flow 1, you could click on the Input for the Timestamp under “Dot Matrix” to see the “Alarm” message scroll on the SenseHat LED. You can see the screenshots for these dashboards in previous blogs.
We can continue running the next sample that will reuse this Node Red deployment. If the Node Red Deployment is no longer required, we can delete it as follows:
cd ~/microshift/raspberry-pi/nodered
oc delete -f noderedpv.yaml -f noderedpvc.yaml -f nodered3.yaml -f noderedroute.yaml -n nodered
oc project default
oc delete project nodered
3. TensorFlow Lite Python object detection example in MicroShift with SenseHat and Node Red
This example requires the same Node Red setup as in the previous Sample 2.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/object-detection
We will build the image for object detection send pictures and web socket chat messages to Node Red when a person is detected using a pod in microshift.
# Use buildah or podman to build the image for object detection
podman build -t docker.io/karve/object-detection-raspberrypi4 .
#buildah bud -t docker.io/karve/object-detection-raspberrypi4 .
podman push docker.io/karve/object-detection-raspberrypi4:latest
Update the env WebSocketURL and ImageUploadURL as shown below. Also update the hostAliases in object-detection.yaml to point to your raspberry pi 4 ip address (192.168.1.227 shown below).
env:
- name: WebSocketURL
value: "ws://nodered-svc-nodered.cluster.local/ws/chat"
- name: ImageUploadURL
value: http://nodered-svc-nodered.cluster.local/upload
hostAliases:
- hostnames:
- nodered-svc-nodered.cluster.local
ip: 192.168.1.227
Create the deployment
oc project default
oc apply -f object-detection.yaml
oc -n default wait deployment object-detection-deployment --for condition=Available --timeout=300s
We will see pictures being sent to Node Red when a person is detected at http://nodered-svc-nodered.cluster.local/#flow/3e30dc50ae28f61f and chat messages at http://nodered-svc-nodered.cluster.local/chat. When we are done testing, we can delete the deployment
oc delete -f object-detection.yaml
4. Running a Virtual Machine Instance on MicroShift
Find the latest version of the KubeVirt Operator.
LATEST=$(curl -L https://storage.googleapis.com/kubevirt-prow/devel/nightly/release/kubevirt/kubevirt/latest-arm64)
echo $LATEST
I used the following version:
LATEST=20220523 # If the latest version does not work
oc apply -f https://storage.googleapis.com/kubevirt-prow/devel/nightly/release/kubevirt/kubevirt/${LATEST}/kubevirt-operator-arm64.yaml
oc apply -f https://storage.googleapis.com/kubevirt-prow/devel/nightly/release/kubevirt/kubevirt/${LATEST}/kubevirt-cr-arm64.yaml
oc adm policy add-scc-to-user privileged -n kubevirt -z kubevirt-operator
# The .status.phase will show Deploying multiple times and finally Deployed
oc get kubevirt.kubevirt.io/kubevirt -n kubevirt -o=jsonpath="{.status.phase}" -w # Ctrl-C to break
oc -n kubevirt wait kv kubevirt --for condition=Available --timeout=300s
oc get pods -n kubevirt
We can build the OKD Web Console (Codename: “bridge”) from the source as mentioned in Part 9. We will run the “bridge” as a container image that we run within MicroShift.
cd /root/microshift/raspberry-pi/console
oc create serviceaccount console -n kube-system
oc create clusterrolebinding console --clusterrole=cluster-admin --serviceaccount=kube-system:console -n kube-system
sleep 5
oc get serviceaccount console --namespace=kube-system -o jsonpath='{.secrets[0].name}'
oc get serviceaccount console --namespace=kube-system -o jsonpath='{.secrets[1].name}'
Replace BRIDGE_K8S_MODE_OFF_CLUSTER_ENDPOINT value https://192.168.1.209:6443 with your raspberry pi 4's ip address, and secretRef token with the console-token-* from above two secret names for BRIDGE_K8S_AUTH_BEARER_TOKEN in okd-web-console-install.yaml. Then apply/create the okd-web-console-install.yaml.
oc apply -f okd-web-console-install.yaml
oc expose svc console-np-service -n kube-system
oc get routes -n kube-system
oc -n kube-system wait deployment console-deployment --for condition=Available --timeout=300s
oc logs deployment/console-deployment -f -n kube-system
Add the Raspberry Pi IP address to /etc/hosts on your Macbook Pro to resolve console-np-service-kube-system.cluster.local. Now you can access the OKD Web Console from your Laptop http://console-np-service-kube-system.cluster.local/. If you see a blank page, you probably have the value of BRIDGE_K8S_MODE_OFF_CLUSTER_ENDPOINT set incorrectly.
We can optionally preload the fedora image into crio (if using the all-in-one containerized approach, this needs to be run within the microshift pod running in podman)
crictl pull quay.io/kubevirt/fedora-cloud-container-disk-demo:20210811_9fec1f849-arm64
Now let’s create a Fedora Virtual Machine Instance using the vmi-fedora.yaml.
cd /root/microshift/raspberry-pi/vmi
oc apply -f vmi-fedora.yaml
watch oc get vmi,pods
The output for the virtualmachineinstance PHASE goes from “Scheduling” to “Scheduled” to “Running” after the virt-launcher-vmi-fedora pod STATUS goes from “Init” to “Running”. Note down the ip address of the vmi-fedora Virtual Machine Instance. Directly connect to the VMI from the Raspberry Pi 4 with fedora as the password. Note that it will take another minute after the VMI goes to Running state to ssh to the instance.
Output:
[root@rpi vmi]# oc get vmi
NAME AGE PHASE IP NODENAME READY
vmi-fedora 6m20s Running 10.85.0.28 rpi.example.com True
[root@rpi vmi]# ssh -o StrictHostKeyChecking=no fedora@10.85.0.28 ping -c 2 google.com
Warning: Permanently added '10.85.0.28' (ED25519) to the list of known hosts.
fedora@10.85.0.28's password:
PING google.com (142.250.80.78) 56(84) bytes of data.
64 bytes from lga34s35-in-f14.1e100.net (142.250.80.78): icmp_seq=1 ttl=116 time=5.49 ms
64 bytes from lga34s35-in-f14.1e100.net (142.250.80.78): icmp_seq=2 ttl=116 time=6.25 ms
--- google.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1001ms
rtt min/avg/max/mdev = 5.489/5.869/6.250/0.380 ms
Alternatively, a second way is to create a Pod to run the ssh client and connect to the Fedora VM from this pod. Let’s create that openssh-client pod:
oc run alpine --privileged --rm -ti --image=alpine -- /bin/sh
apk update && apk add --no-cache openssh-client
or
oc run sshclient --privileged --rm -ti --image=karve/alpine-sshclient:arm64 -- /bin/sh
#oc attach sshclient -c sshclient -i -t
Then, ssh to the Fedora VMI from this openssh-client container.
Output:
[root@rpi vmi]# oc run sshclient --privileged --rm -ti --image=karve/alpine-sshclient:arm64 -- /bin/sh
If you don't see a command prompt, try pressing enter.
/ # ssh -o StrictHostKeyChecking=no fedora@10.85.0.28 "bash -c \"ping -c 2 google.com\""
Warning: Permanently added '10.85.0.28' (ED25519) to the list of known hosts.
fedora@10.85.0.28's password:
PING google.com (142.250.80.78) 56(84) bytes of data.
64 bytes from lga34s35-in-f14.1e100.net (142.250.80.78): icmp_seq=1 ttl=116 time=5.70 ms
64 bytes from lga34s35-in-f14.1e100.net (142.250.80.78): icmp_seq=2 ttl=116 time=6.44 ms
--- google.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1002ms
rtt min/avg/max/mdev = 5.700/6.068/6.436/0.368 ms
/ # exit
Session ended, resume using 'oc attach sshclient -c sshclient -i -t' command when the pod is running
pod "sshclient" deleted
A third way to connect to the VM is to use the virtctl console. You can compile your own virtctl as was described in Part 9. To simplify, we copy virtctl arm64 binary from prebuilt container image to /usr/local/bin on the Raspberry Pi 4 and connect to the VMI using “virtctl console” command.
id=$(podman create docker.io/karve/kubevirt:arm64)
podman cp $id:_out/cmd/virtctl/virtctl /usr/local/bin
podman rm -v $id
virtctl console vmi-fedora
Output:
[root@rpi vmi]# id=$(podman create docker.io/karve/kubevirt:arm64)
Trying to pull docker.io/karve/kubevirt:arm64...
Getting image source signatures
Copying blob 7065f6098427 done
Copying config 1c7a5aa443 done
Writing manifest to image destination
Storing signatures
[root@rpi vmi]# podman cp $id:_out/cmd/virtctl/virtctl /usr/local/bin
[root@rpi vmi]# podman rm -v $id
c34b3943d54c37a1d72ab61d68365bba90f8bb792cebac68b803bfc7e1f27aa4
[root@rpi vmi]# virtctl console vmi-fedora
Successfully connected to vmi-fedora console. The escape sequence is ^]
vmi-fedora login: fedora
Password:
[systemd]
Failed Units: 1
dnf-makecache.service
[fedora@vmi-fedora ~]$ ping -c 2 google.com
PING google.com (142.250.72.110) 56(84) bytes of data.
64 bytes from lga34s32-in-f14.1e100.net (142.250.72.110): icmp_seq=1 ttl=116 time=5.41 ms
64 bytes from lga34s32-in-f14.1e100.net (142.250.72.110): icmp_seq=2 ttl=116 time=3.98 ms
--- google.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1002ms
rtt min/avg/max/mdev = 3.979/4.695/5.411/0.716 ms
[fedora@vmi-fedora ~]$ # Ctrl-] to detach
[root@rpi vmi]#
When done, we can delete the VMI
oc delete -f vmi-fedora.yaml
We can run other VM and VMI samples for alpine, cirros and fedora images as in Part 9. When done, you may delete kubevirt operator
oc delete -f https://storage.googleapis.com/kubevirt-prow/devel/nightly/release/kubevirt/kubevirt/${LATEST}/kubevirt-cr-arm64.yaml
oc delete -f https://storage.googleapis.com/kubevirt-prow/devel/nightly/release/kubevirt/kubevirt/${LATEST}/kubevirt-operator-arm64.yaml
5. Run a jupyter notebook sample for license plate recognition (RPi with 8GB RAM)
We will run the sample described at the Red Hat OpenShift Data Science Workshop License plate recognition. The Dockerfile uses the arm64 Jupyter Notebook base image: scipy-notebook. Since we do not have a tensorflow arm64 image, we install it as described at Qengineering. The notebook.yaml downloads the licence-plate-workshop sample in an initContainer.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/tensorflow-notebook
oc apply -f notebook.yaml
oc -n default wait pod notebook --for condition=Ready --timeout=600s
oc get routes
Output:
[root@rpi tensorflow-notebook]# oc get routes
NAME HOST/PORT PATH SERVICES PORT TERMINATION WILDCARD
flask-route flask-route-default.cluster.local notebook-svc 5000 None
notebook-route notebook-route-default.cluster.local notebook-svc 5001 None
The image is large, it may take a while for image to be downloaded:
[root@rpi tensorflow-notebook]# crictl images | grep tensorflow-notebook
docker.io/karve/tensorflow-notebook arm64 c8da62870fec2 4.73GB
If running in the all-in-one microshift container, you need to run the command within the container
[root@rpi tensorflow-notebook]# # podman exec -it microshift crictl images | grep tensorflow-notebook # All in one
Add the ipaddress of the Raspberry Pi 4 device for notebook-route-default.cluster.local to /etc/hosts on your Laptop and browse to http://notebook-route-default.cluster.local/tree?. Login with the default password mysecretpassword. Go the work folder and select and run the License-plate-recognition notebook at http://notebook-route-default.cluster.local/notebooks/work/02_Licence-plate-recognition.ipynb
We can also run it as an application and test it using the corresponding notebooks. Run the http://notebook-route-default.cluster.local/notebooks/work/03_LPR_run_application.ipynb
Wait for the following to appear.
Instructions for updating:
non-resource variables are not supported in the long term
Model Loaded successfully...
Model Loaded successfully...
[INFO] Model loaded successfully...
[INFO] Labels loaded successfully...
Then run http://notebook-route-default.cluster.local/notebooks/work/04_LPR_test_application.ipynb
We can experiment with a custom image. Let’s download the image to the pod and run the cells again with the new image and check the prediction.
oc exec -it notebook -- bash -c "wget \"https://unsplash.com/photos/MgfKoRdI948/download?force=true&ixid=MnwxMjA3fDB8MXxhbGx8fHx8fHx8fHwxNjUyNDY4Mjcz\" -O /tmp/3183KND.jpg"
Then, run the http://notebook-route-default.cluster.local/notebooks/work/05_Send_image.ipynb
Add the cell with the following code:
my_image = 'https://unsplash.com/photos/MgfKoRdI948/download?force=true&ixid=MnwxMjA3fDB8MXxhbGx8fHx8fHx8fHwxNjUyNDY4Mjcz'
from PIL import Image
import requests
from io import BytesIO
response = requests.get(my_image)
img = BytesIO(response.content).read()
import base64
import requests
from json import dumps
encoded_image = base64.b64encode(img).decode('utf-8')
content = {"image": encoded_image}
json_data = dumps(content)
headers = {"Content-Type" : "application/json"}
r = requests.post(my_route + '/predictions', data=json_data, headers=headers)
print(r.content)
from IPython.display import Image
from IPython.core.display import HTML
Image(url=my_image)
When we are done working with the license plate recognition sample notebook, we can delete it as follows:
oc delete -f notebook.yaml
6. Run a jupyter notebook sample for object detection
We will run the sample described at the Red Hat OpenShift Data Science Workshop Object Detection. We use the same container image as in previous Sample 5, the only change is to download the object detection sample in object-detection-rest.yaml from object-detection-rest.git.
cd ~
git clone https://github.com/thinkahead/microshift.git
cd ~/microshift/raspberry-pi/tensorflow-notebook
oc apply -f object-detection-rest.yaml
oc -n default wait pod notebook --for condition=Ready --timeout=300s
oc get routes
Output will look the same as in Sample 5; we use the same service and route names.
[root@rpi tensorflow-notebook]# oc apply -f object-detection-rest.yaml
pod/notebook created
service/flask-svc created
service/notebook-svc created
route.route.openshift.io/notebook-route created
route.route.openshift.io/flask-route created
[root@rpi tensorflow-notebook]# oc get routes
NAME HOST/PORT PATH SERVICES PORT TERMINATION WILDCARD
flask-route flask-route-default.cluster.local notebook-svc 5000 None
notebook-route notebook-route-default.cluster.local notebook-svc 5001 None
Login at http://notebook-route-default.cluster.local/tree/work with the default password mysecretpassword. We can run the 1_explore.ipynb that will download twodogs.jpg and use a pre-trained model to identify objects in images. In the next notebooks (2_predict.ipynb, 3_run_flask.ipynb, and 4_test_flask.ipynb), this model is wrapped in a flask app that can be used as part of a larger application.
In 4_test_flask.ipynb, replace the my_route as follows:
my_route = 'http://flask-svc:5000'
We can also test by downloading custom images, for example from Dogs Best Life.
oc exec -it notebook -- bash -c "wget https://dogsbestlife.com/wp-content/uploads/2016/05/two-dogs-same-litter-min.jpeg -O /home/jovyan/work/two-dogs-same-litter-min.jpeg"
In 4_test_flask.ipynb, replace the my_image and run the notebook.
my_image = 'two-dogs-same-litter-min.jpeg'
When we are done working with the object detection sample notebook, we can delete it as follows:
oc delete -f object-detection-rest.yaml
7. Tutorial Notebooks from tensorflow.org
We can run the tutorials from https://www.tensorflow.org/tutorials using the tutorials.yaml. We use the same container image as in previous Sample 5, the only change is that it pulls notebooks from https://github.com/tensorflow/docs.git. You will need to make a few minor changes to use the /tmp to download temporary file and cache folders used in the notebooks to avoid permission denied errors because the local directory is not writable.
[root@rpi vmi]# cd ~/microshift/raspberry-pi/tensorflow-notebook
[root@rpi tensorflow-notebook]# oc apply -f tutorials.yaml
pod/notebook created
service/flask-svc created
service/notebook-svc created
route.route.openshift.io/notebook-route created
route.route.openshift.io/flask-route created
[root@rpi tensorflow-notebook]# oc get routes notebook-route
NAME HOST/PORT PATH SERVICES PORT TERMINATION WILDCARD
notebook-route notebook-route-default.cluster.local notebook-svc 5001 None
1. Tensorflow 2 quickstart beginner http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/quickstart/beginner.ipynb
2. TensorFlow 2 quickstart for experts http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/quickstart/advanced.ipynb
3. Segmentation http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/images/segmentation.ipynb
4. Classification of flowers that shows overfitting, data augmentation for generating additional training data http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/images/classification.ipynb
5. Audio recognition: Recognizing keywords http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/audio/simple_audio.ipynb
6. Time series forecasting - It builds a few different styles of models including Convolutional and Recurrent Neural Networks (CNNs and RNNs) http://notebook-route-default.cluster.local/notebooks/work/site/en/tutorials/structured_data/time_series.ipynb
Cleanup MicroShift
We can use the cleanup.sh script available on github to cleanup the pods and images. If you already cloned the microshift repo from github, you have the script in the ~/microshift/hack directory.
cd ~/microshift/hack
./cleanup.sh
Containerized MicroShift on Manjaro (64 bit)
We can run MicroShift within containers in two ways:
- MicroShift Containerized – The MicroShift binary runs in a Podman container, CRI-O Systemd service runs directly on the host and data is stored in a podman volume
- MicroShift Containerized All-In-One – The MicroShift binary and CRI-O service run within a container and data is stored in a podman volume, microshift-data. This should be used for “Testing and Development” only
Microshift Containerized
If you did not already install podman, you can do it now.
pacman -S podman
We will use a new microshift.service that runs microshift in a pod using the prebuilt image and uses a podman volume. Rest of the pods run using crio on the host.
cat << EOF > /usr/lib/systemd/system/microshift.service
[Unit]
Description=MicroShift Containerized
Documentation=man:podman-generate-systemd(1)
Wants=network-online.target crio.service
After=network-online.target crio.service
RequiresMountsFor=%t/containers
[Service]
Environment=PODMAN_SYSTEMD_UNIT=%n
Restart=on-failure
TimeoutStopSec=70
ExecStartPre=/usr/bin/mkdir -p /var/lib/kubelet ; /usr/bin/mkdir -p /var/hpvolumes
ExecStartPre=/bin/rm -f %t/%n.ctr-id
ExecStart=/bin/podman run \
--cidfile=%t/%n.ctr-id \
--cgroups=no-conmon \
--rm \
--replace \
--sdnotify=container \
--label io.containers.autoupdate=registry \
--network=host \
--privileged \
-d \
--name microshift \
-v /var/hpvolumes:/var/hpvolumes:z,rw,rshared \
-v /var/run/crio/crio.sock:/var/run/crio/crio.sock:rw,rshared \
-v microshift-data:/var/lib/microshift:rw,rshared \
-v /var/lib/kubelet:/var/lib/kubelet:z,rw,rshared \
-v /var/log:/var/log \
-v /etc:/etc quay.io/microshift/microshift:latest
ExecStop=/bin/podman stop --ignore --cidfile=%t/%n.ctr-id
ExecStopPost=/bin/podman rm -f --ignore --cidfile=%t/%n.ctr-id
Type=notify
NotifyAccess=all
[Install]
WantedBy=multi-user.target default.target
EOF
systemctl daemon-reload
systemctl enable --now crio microshift
podman ps -a
podman volume inspect microshift-data # Get the Mountpoint where kubeconfig is located
export KUBECONFIG=/var/lib/containers/storage/volumes/microshift-data/_data/resources/kubeadmin/kubeconfig
watch "oc get nodes;oc get pods -A;crictl pods;crictl images;podman ps"
Output:
NAME STATUS ROLES AGE VERSION
rpi.example.com Ready <none> 2m45s v1.21.0
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system kube-flannel-ds-pgkmc 1/1 Running 0 2m45s
kubevirt-hostpath-provisioner kubevirt-hostpath-provisioner-vf9ch 1/1 Running 0 2m35s
openshift-dns dns-default-hzqff 2/2 Running 0 2m45s
openshift-dns node-resolver-gct24 1/1 Running 0 2m45s
openshift-ingress router-default-85bcfdd948-bdb7p 1/1 Running 0 2m48s
openshift-service-ca service-ca-7764c85869-ss4jx 1/1 Running 0 2m50s
POD ID CREATED STATE NAME NAMESPACE ATTEMPT
RUNTIME
cfb2ca0152657 28 seconds ago Ready dns-default-hzqff openshift-dns 0
(default)
70a054df7ac5f About a minute ago Ready router-default-85bcfdd948-bdb7p openshift-ingress 0
(default)
2fa32698761df 2 minutes ago Ready service-ca-7764c85869-ss4jx openshift-service-ca 0
(default)
4d5ec69042155 2 minutes ago Ready kubevirt-hostpath-provisioner-vf9ch kubevirt-hostpath-provisioner 0
(default)
3a3dfcafc14df 2 minutes ago Ready kube-flannel-ds-pgkmc kube-system 0
(default)
be5fcdba15e33 2 minutes ago Ready node-resolver-gct24 openshift-dns 0
(default)
IMAGE TAG IMAGE ID SIZE
k8s.gcr.io/pause 3.6 7d46a07936af9 492kB
quay.io/microshift/cli 4.8.0-0.okd-2021-10-10-030117 33a276ba2a973 205MB
quay.io/microshift/coredns 4.8.0-0.okd-2021-10-10-030117 67a95c8f15902 265MB
quay.io/microshift/flannel-cni 4.8.0-0.okd-2021-10-10-030117 0e66d6f50c694 8.78MB
quay.io/microshift/flannel 4.8.0-0.okd-2021-10-10-030117 85fc911ceba5a 68.1MB
quay.io/microshift/haproxy-router 4.8.0-0.okd-2021-10-10-030117 37292c44812e7 225MB
quay.io/microshift/hostpath-provisioner 4.8.0-0.okd-2021-10-10-030117 fdef3dc1264ad 39.3MB
quay.io/microshift/kube-rbac-proxy 4.8.0-0.okd-2021-10-10-030117 7f149e453e908 41.5MB
quay.io/microshift/microshift latest bdccb7de6c282 406MB
quay.io/microshift/service-ca-operator 4.8.0-0.okd-2021-10-10-030117 0d3ab44356260 276MB
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
37391110a83f quay.io/microshift/microshift:latest run 4 minutes ago Up 4 minutes ago microshift
Now that microshift is started, we can run the samples shown earlier.
After we are done, we can stop microshift
systemctl stop microshift
podman volume rm microshift-data
Alternatively, delete the microshift container. The --rm we used in the podman run will delete the container when we stop it.
podman stop microshift && podman volume rm microshift-data
After it is stopped, we can run the cleanup.sh to delete the pods and images from crio.
MicroShift Containerized All-In-One
Let’s stop the crio on the host, we will be creating an all-in-one container in podman that will run crio within the container.
systemctl stop crio
systemctl disable crio
mkdir /var/hpvolumes
We will run the all-in-one microshift in podman using prebuilt images (replace the image in the podman run command below with the latest image).
podman volume rm microshift-data;podman volume create microshift-data
podman run -d --rm --name microshift -h microshift.example.com --privileged -v /lib/modules:/lib/modules -v microshift-data:/var/lib -v /var/hpvolumes:/var/hpvolumes -p 6443:6443 -p 8080:8080 -p 80:80 quay.io/microshift/microshift-aio:4.8.0-0.microshift-2022-04-20-182108-linux-nft-arm64
Now that you know the podman command to start the microshift all-in-one, you may alternatively use the following microshift service.
wget https://raw.githubusercontent.com/thinkahead/microshift/main/packaging/systemd/microshift-aio.service -O /usr/lib/systemd/system/microshift.service
# Add the “-p 80:80” after the “-p 6443:6443” so we can expose the applications
# Add the “-h microshift.example.com”
or
cat << EOF > /usr/lib/systemd/system/microshift.service
[Unit]
Description=MicroShift all-in-one
Documentation=man:podman-generate-systemd(1)
Wants=network-online.target
After=network-online.target
RequiresMountsFor=%t/containers
[Service]
Environment=PODMAN_SYSTEMD_UNIT=%n
Restart=on-failure
TimeoutStopSec=70
ExecStartPre=/bin/rm -f %t/%n.ctr-id
ExecStart=/usr/bin/podman run --cidfile=%t/%n.ctr-id --sdnotify=conmon --cgroups=no-conmon --rm --replace -d --name microshift -h microshift.example.com --privileged -v /sys/fs/cgroup:/sys/fs/cgroup:ro -v microshift-data:/var/lib -v /var/hpvolumes:/var/hpvolumes -v /lib/modules:/lib/modules --label io.containers.autoupdate=registry -p 6443:6443 -p 80:80 quay.io/microshift/microshift-aio:latest
ExecStop=/usr/bin/podman stop --ignore --cidfile=%t/%n.ctr-id
ExecStopPost=/usr/bin/podman rm -f --ignore --cidfile=%t/%n.ctr-id
Type=notify
NotifyAccess=all
[Install]
WantedBy=multi-user.target default.target
EOF
Then run:
systemctl daemon-reload
systemctl start microshift
On the host Raspberry Pi 4, we set KUBECONFIG to point to the kubeconfig on the data volume at the Mountpoint from above.
export KUBECONFIG=/var/lib/containers/storage/volumes/microshift-data/_data/microshift/resources/kubeadmin/kubeconfig
# crio on host is stopped, so we do not run crictl commands on host
watch "oc get nodes;oc get pods -A;podman exec -it microshift crictl ps -a"
Output:
NAME STATUS ROLES AGE VERSION
1a0e4de49f0c Ready <none> 3m24s v1.21.0
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system kube-flannel-ds-x2gnq 1/1 Running 0 3m23s
kubevirt-hostpath-provisioner kubevirt-hostpath-provisioner-72bj5 1/1 Running 0 2m33s
openshift-dns dns-default-rnprr 2/2 Running 0 3m24s
openshift-dns node-resolver-cmjqw 1/1 Running 0 3m24s
openshift-ingress router-default-85bcfdd948-d2fkn 1/1 Running 0 3m27s
openshift-service-ca service-ca-7764c85869-vcv55 1/1 Running 0 3m29s
CONTAINER IMAGE CREATED
STATE NAME ATTEMPT POD IDD
66e45f02e2a40 quay.io/microshift/kube-rbac-proxy@sha256:2b5f44b11bab4c10138ce526439b43d62a890c3a02d42893ad02e2b3adb38703 48 seconds ago
Running kube-rbac-proxy 0 188183a0488679
6c0dcd12de6a0 quay.io/microshift/coredns@sha256:07e5397247e6f4739727591f00a066623af9ca7216203a5e82e0db2fb24514a3 55 seconds ago
Running dns 0 188183a0488675
bf2677b89d554 quay.io/microshift/haproxy-router@sha256:706a43c785337b0f29aef049ae46fdd65dcb2112f4a1e73aaf0139f70b14c6b5 56 seconds ago
Running router 0 e18109aeeb3edb
28ab2b3973ad3 quay.io/microshift/hostpath-provisioner@sha256:cb0c1cc60c1ba90efd558b094ba1dee3a75e96b76e3065565b60a07e4797c04c About a minute ag
o Running kubevirt-hostpath-provisioner 0 25c8df5729dbd7
fac5960bf4739 quay.io/microshift/service-ca-operator@sha256:1a8e53c8a67922d4357c46e5be7870909bb3dd1e6bea52cfaf06524c300b84e8 2 minutes ago
Running service-ca-controller 0 4b3c3fab45afcb
dac23f94cfd67 85fc911ceba5a5a5e43a7c613738b2d6c0a14dad541b1577cdc6f921c16f5b75 2 minutes ago
Running kube-flannel 0 67b1085d9bec9b
8f5dc2a9657b5 quay.io/microshift/cli@sha256:1848138e5be66753863c98b86c274bd7fb8572fe0da6f7156f1644187e4cfb84 2 minutes ago
Running dns-node-resolver 0 f3da680ae7071
5795218a71896 quay.io/microshift/flannel@sha256:13777a318497ae35593bb73499a0e2ff4cb7eda74f59c1ba7c3e88c717cbaab9 2 minutes ago
Exited install-cni 0 67b1085d9bec9
098607fa50de6 quay.io/microshift/flannel-cni@sha256:39f81dd125398ce5e679322286344a4c13dded73ea0bf4f397e5d1929b43d033 3 minutes ago
Exited install-cni-bin 0 67b1085d9bec9
The crio service is stopped on the Raspberry Pi, so crictl command will not work directly on the Pi. The crictl commands will work within the microshift container in podman as shown in the watch command above.
Now, we can run the samples shown earlier. To run the Virtual Machine examples in the all-in-one MicroShift, we need to execute the mount with --make-shared as follows in the microshift container to prevent the “Error: path "/var/run/kubevirt" is mounted on "/" but it is not a shared mount” event from virt-handler.
podman exec -it microshift mount --make-shared /
We may also preload the virtual machine images using "crictl pull".
podman exec -it microshift crictl pull quay.io/kubevirt/fedora-cloud-container-disk-demo:20210811_9fec1f849-arm64
For the Virtual Machine Instance Sample 4, we can connect to the vmi-fedora by exposing the ssh port for the Virtual Machine Instance as a NodePort Service after the instance is started. This NodePort is within the all-in-one pod that is running in podman. The ip address of the all-in-one microshift podman container is 10.88.0.2. We expose the target port 22 on the VM as a service on port 22 that is in turn exposed on the microshift container with allocated port 30339 as seen below. We run and exec into a new pod called ssh-proxy, install the openssh-client on the ssh-proxy and ssh to the port 30339 on the all-in-one microshift container. This takes us to the VMI port 22 as shown below:
[root@rpi vmi]# oc get vmi,pods
NAME AGE PHASE IP NODENAME READY
virtualmachineinstance.kubevirt.io/vmi-fedora 3m17s Running 10.42.0.13 1a0e4de49f0c True
NAME READY STATUS RESTARTS AGE
pod/virt-launcher-vmi-fedora-kp8zz 2/2 Running 0 3m16s
[root@rpi vmi]# virtctl expose vmi vmi-fedora --port=22 --target-port=22 --name=vmi-fedora-ssh --type=NodePort
Service vmi-fedora-ssh successfully exposed for vmi vmi-fedora
[root@rpi vmi]# oc get svc vmi-fedora-ssh
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
vmi-fedora-ssh NodePort 10.43.248.91 <none> 22:32462/TCP 12s
[root@rpi vmi]# podman inspect --format "{{.NetworkSettings.IPAddress}}" microshift
10.88.0.4
[root@rpi vmi]# oc run -i --tty ssh-proxy --rm --image=karve/alpine-sshclient:arm64 --restart=Never -- /bin/sh -c "ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null fedora@10.88.0.4 -p 32462"
If you don't see a command prompt, try pressing enter.
[fedora@vmi-fedora ~]$ sudo dnf install -y qemu-guest-agent >/dev/null
[fedora@vmi-fedora ~]$ sudo systemctl enable --now qemu-guest-agent
[fedora@vmi-fedora ~]$ ping -c 2 google.com
PING google.com (142.251.40.206) 56(84) bytes of data.
64 bytes from lga34s38-in-f14.1e100.net (142.251.40.206): icmp_seq=1 ttl=115 time=6.14 ms
64 bytes from lga34s38-in-f14.1e100.net (142.251.40.206): icmp_seq=2 ttl=115 time=6.08 ms
--- google.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1002ms
rtt min/avg/max/mdev = 6.083/6.111/6.139/0.028 ms
[fedora@vmi-fedora ~]$ exit
logout
Connection to 10.88.0.4 closed.
pod "ssh-proxy" deleted
The QEMU guest agent that we installed is a daemon that runs on the virtual machine and passes information to the host about the virtual machine, users, file systems, and secondary networks.
After we are done, we can delete the all-in-one microshift container.
podman rm -f microshift && podman volume rm microshift-data
or if started using systemd, then
systemctl stop microshift
Error Messages in MicroShift Logs
Message: failed to fetch hugetlb info
The following journalctl command will continuously show warning messages “failed to fetch hugetlb info”. The default kernel for the Raspberry Pi OS does not support HugeTLB hugepages.
journalctl -u microshift -f
Output:
...
May 23 11:49:20 rpi.example.com microshift[3112]: W0523 11:49:20.347604 3112 container.go:586] Failed to update stats for container "/system.slice/crio-00c0b63eeee509979e7652cca8b91a1e9e900c1989b7fe54f6a05f6591de0108.scope": error while statting cgroup v2: [open /sys/kernel/mm/hugepages: no such file or directory
May 23 11:49:20 rpi.example.com microshift[3112]: failed to fetch hugetlb info
...
To remove these messages, we can recompile the microshift binary using the changes from hugetlb.go.
pacman -S pkg-config
# Install golang
wget https://golang.org/dl/go1.17.2.linux-arm64.tar.gz
rm -rf /usr/local/go && tar -C /usr/local -xzf go1.17.2.linux-arm64.tar.gz
rm -f go1.17.2.linux-arm64.tar.gz
export PATH=$PATH:/usr/local/go/bin
export GOPATH=/root/go
cat << EOF >> /root/.bashrc
export PATH=$PATH:/usr/local/go/bin
export GOPATH=/root/go
EOF
mkdir $GOPATH
git clone https://github.com/thinkahead/microshift.git
cd microshift
Edit the file vendor/github.com/opencontainers/runc/libcontainer/cgroups/fs2/hugetlb.go and remove the return on err.
func statHugeTlb(dirPath string, stats *cgroups.Stats) error {
hugePageSizes, _ := cgroups.GetHugePageSize()
//hugePageSizes, err := cgroups.GetHugePageSize()
//if err != nil {
// return errors.Wrap(err, "failed to fetch hugetlb info")
//}
hugetlbStats := cgroups.HugetlbStats{}
Build and replace the microshift binary. Restart MicroShift.
make
mv microshift /usr/local/bin/.
systemctl restart microshift
Conclusion
In this Part 18, we saw multiple options to run MicroShift on the Raspberry Pi 4 with the Manjaro (64 bit). We used dynamic persistent volumes to install InfluxDB/Telegraf/Grafana with a dashboard to show SenseHat sensor data. We ran samples that used the Sense Hat/USB camera and worked with a sample that sent the pictures and web socket messages to Node Red when a person was detected. We installed the OKD Web Console and saw how to connect to a Virtual Machine Instance using KubeVirt on MicroShift with Manjaro. Finally, we saw how to run jupyter notebooks with the license plate recognition, object detection, image segmentation, image classification and audio keyword recognition. We will return to Manjaro in Part 24 where we will build and install the Kata Containers Runtime. In the next Part 19, we will work on Kali Linux and in Part 20 with ArchLinux.
Hope you have enjoyed the article. Share your thoughts in the comments or engage in the conversation with me on Twitter @aakarve. I look forward to hearing about your use of MicroShift on ARM devices and if you would like to see something covered in more detail.
References