Skip to content

Commit 765912e

Browse files
ArangoGutierrezArangoGutierrez
authored
Merge pull request #315 from ArangoGutierrez/containerd20
Enable containerd 2.0+ installation
2 parents b6014e5 + 0017e76 commit 765912e

File tree

4 files changed

+486
-27
lines changed

4 files changed

+486
-27
lines changed

‎examples/aws_kubeadm.yaml‎

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
apiVersion: holodeck.nvidia.com/v1alpha1
2+
kind: Environment
3+
metadata:
4+
name: aws_kubeadm_example
5+
description: "end-to-end test infrastructure"
6+
spec:
7+
provider: aws
8+
auth:
9+
keyName: <your key name here>
10+
privateKey: <your key path here>
11+
instance:
12+
type: g4dn.xlarge
13+
region: us-west-1
14+
ingressIpRanges:
15+
- <youip/32>
16+
image:
17+
architecture: amd64
18+
nvidiaDriver:
19+
install: true
20+
nvidiaContainerToolkit:
21+
install: true
22+
containerRuntime:
23+
install: true
24+
name: containerd
25+
kubernetes:
26+
install: true
27+
installer: kubeadm

‎pkg/provisioner/templates/container-toolkit.go‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -34,7 +34,7 @@ curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dear
3434
&& \
3535
sudo apt-get update
3636
37-
sudo apt-get install -y nvidia-container-toolkit
37+
install_packages_with_retry nvidia-container-toolkit
3838
3939
# Configure container runtime
4040
sudo nvidia-ctk runtime configure --runtime={{.ContainerRuntime}} --set-as-default --enable-cdi={{.EnableCDI}}

‎pkg/provisioner/templates/containerd.go‎

Lines changed: 283 additions & 23 deletions
Original file line numberDiff line numberDiff line change
@@ -28,33 +28,293 @@ import (
2828
const containerdTemplate = `
2929
: ${CONTAINERD_VERSION:={{.Version}}}
3030
31+
# Check system requirements
32+
echo "Checking system requirements..."
33+
34+
# Check for systemd
35+
if ! command -v systemctl &> /dev/null; then
36+
echo "Error: systemd is required but not installed"
37+
exit 1
38+
fi
39+
40+
# Check and load required kernel modules
41+
echo "Checking and loading required kernel modules..."
42+
REQUIRED_MODULES="overlay br_netfilter"
43+
for module in $REQUIRED_MODULES; do
44+
if ! lsmod | grep -q "^${module}"; then
45+
echo "Loading ${module} module..."
46+
if ! sudo modprobe ${module}; then
47+
echo "Error: Failed to load ${module} module"
48+
exit 1
49+
fi
50+
fi
51+
done
52+
53+
# Ensure modules are loaded at boot
54+
for module in $REQUIRED_MODULES; do
55+
if [ ! -f "/etc/modules-load.d/${module}.conf" ]; then
56+
echo "${module}" | sudo tee "/etc/modules-load.d/${module}.conf" > /dev/null
57+
fi
58+
done
59+
60+
# Check and configure sysctl settings
61+
echo "Configuring sysctl settings..."
62+
SYSCTL_SETTINGS=(
63+
"net.bridge.bridge-nf-call-iptables=1"
64+
"net.bridge.bridge-nf-call-ip6tables=1"
65+
"net.ipv4.ip_forward=1"
66+
)
67+
68+
for setting in "${SYSCTL_SETTINGS[@]}"; do
69+
key=$(echo $setting | cut -d= -f1)
70+
value=$(echo $setting | cut -d= -f2)
71+
if [ "$(sudo sysctl -n $key)" != "$value" ]; then
72+
echo "Setting $key to $value"
73+
sudo sysctl -w $key=$value
74+
echo "$key=$value" | sudo tee -a /etc/sysctl.conf > /dev/null
75+
fi
76+
done
77+
78+
# Apply sysctl settings
79+
sudo sysctl --system
80+
81+
# Check for required commands
82+
REQUIRED_COMMANDS="curl tar systemctl"
83+
for cmd in $REQUIRED_COMMANDS; do
84+
if ! command -v $cmd &> /dev/null; then
85+
echo "Error: Required command '$cmd' not found"
86+
exit 1
87+
fi
88+
done
89+
3190
# Install required packages
91+
echo "Installing required packages..."
3292
with_retry 3 10s sudo apt-get update
3393
install_packages_with_retry ca-certificates curl gnupg -y
3494
sudo install -m 0755 -d /etc/apt/keyrings
35-
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
36-
sudo chmod a+r /etc/apt/keyrings/docker.gpg
37-
38-
# Add the repository to Apt sources:
39-
echo \
40-
"deb [arch="$(dpkg --print-architecture)" signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu \
41-
"$(. /etc/os-release && echo "$VERSION_CODENAME")" stable" | \
42-
sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
43-
with_retry 3 10s sudo apt-get update
44-
45-
# Install containerd
46-
install_packages_with_retry containerd.io=${CONTAINERD_VERSION}-1
47-
48-
# Configure containerd and start service
49-
mkdir -p /etc/containerd
50-
containerd config default | sudo tee /etc/containerd/config.toml
51-
# Set systemd as the cgroup driver
52-
# see https://kubernetes.io/docs/setup/production-environment/container-runtimes/#containerd
53-
sudo sed -i 's/SystemdCgroup \= false/SystemdCgroup \= true/g' /etc/containerd/config.toml
54-
55-
# restart containerd
56-
sudo systemctl restart containerd
57-
sudo systemctl enable containerd
95+
96+
# Check if CONTAINERD_VERSION is empty, if so fetch the latest stable version
97+
if [ -z "$CONTAINERD_VERSION" ]; then
98+
echo "Fetching latest stable containerd version..."
99+
CONTAINERD_VERSION=$(curl -fsSL https://api.github.com/repos/containerd/containerd/releases/latest | grep '"tag_name":' | cut -d '"' -f 4 | sed 's/v//')
100+
101+
if [ -z "$CONTAINERD_VERSION" ]; then
102+
echo "Failed to fetch latest Containerd version. Exiting."
103+
exit 1
104+
fi
105+
fi
106+
echo "Using containerd version: $CONTAINERD_VERSION"
107+
108+
# Determine major version for configuration
109+
MAJOR_VERSION=$(echo $CONTAINERD_VERSION | cut -d. -f1)
110+
111+
# Check architecture
112+
ARCH=$(uname -m)
113+
if [[ "$ARCH" == "x86_64" ]]; then
114+
ARCH="amd64"
115+
elif [[ "$ARCH" == "aarch64" ]]; then
116+
ARCH="arm64"
117+
else
118+
echo "Error: Unsupported architecture: $ARCH"
119+
exit 1
120+
fi
121+
122+
CONTAINERD_TAR="containerd-${CONTAINERD_VERSION}-linux-${ARCH}.tar.gz"
123+
CONTAINERD_URL="https://github.com/containerd/containerd/releases/download/v${CONTAINERD_VERSION}/${CONTAINERD_TAR}"
124+
CONTAINERD_SHA256_URL="${CONTAINERD_URL}.sha256sum"
125+
126+
echo "Downloading and extracting containerd ${CONTAINERD_VERSION} from: $CONTAINERD_URL"
127+
128+
# Create temporary directory for downloads
129+
TMP_DIR=$(mktemp -d)
130+
cd $TMP_DIR
131+
132+
# Download containerd tarball and checksum
133+
if ! curl -fsSL -o ${CONTAINERD_TAR} ${CONTAINERD_URL}; then
134+
echo "Error: Failed to download containerd tarball"
135+
exit 1
136+
fi
137+
138+
if ! curl -fsSL -o containerd_SHA256SUMS ${CONTAINERD_SHA256_URL}; then
139+
echo "Error: Failed to download containerd checksum"
140+
exit 1
141+
fi
142+
143+
# Verify SHA256 checksum
144+
EXPECTED_CHECKSUM=$(cat containerd_SHA256SUMS | awk '{print $1}')
145+
ACTUAL_CHECKSUM=$(sha256sum ${CONTAINERD_TAR} | awk '{print $1}')
146+
147+
if [ "$EXPECTED_CHECKSUM" != "$ACTUAL_CHECKSUM" ]; then
148+
echo "Error: Checksum verification failed for containerd"
149+
echo "Expected: $EXPECTED_CHECKSUM"
150+
echo "Actual: $ACTUAL_CHECKSUM"
151+
exit 1
152+
fi
153+
154+
# Stream directly into tar to avoid saving the archive
155+
if ! cat ${CONTAINERD_TAR} | sudo tar Cxzvf /usr/local -; then
156+
echo "Error: Failed to extract containerd tarball"
157+
exit 1
158+
fi
159+
160+
# Cleanup
161+
cd - > /dev/null
162+
rm -rf $TMP_DIR
163+
164+
echo "Containerd ${CONTAINERD_VERSION} installed successfully."
165+
166+
# Fetch latest stable RUNC version from GitHub
167+
echo "Fetching latest stable runc version..."
168+
RUNC_VERSION=$(curl -fsSL https://api.github.com/repos/opencontainers/runc/releases/latest | grep '"tag_name":' | cut -d '"' -f 4 | sed 's/v//')
169+
170+
if [ -z "$RUNC_VERSION" ]; then
171+
echo "Failed to fetch latest RUNC version. Using default version."
172+
RUNC_VERSION="1.2.6"
173+
fi
174+
175+
RUNC_URL="https://github.com/opencontainers/runc/releases/download/v${RUNC_VERSION}/runc.${ARCH}"
176+
177+
echo "Downloading runc ${RUNC_VERSION} from: $RUNC_URL"
178+
179+
# Download runc binary and checksum
180+
curl -fsSL -o runc.${ARCH} ${RUNC_URL}
181+
182+
sudo install -m 755 runc.${ARCH} /usr/local/sbin/runc
183+
184+
echo "Runc ${RUNC_VERSION} installed successfully."
185+
186+
# Install CNI plugins
187+
CNI_VERSION="1.1.1"
188+
CNI_TAR="cni-plugins-linux-${ARCH}-v${CNI_VERSION}.tgz"
189+
CNI_URL="https://github.com/containernetworking/plugins/releases/download/v${CNI_VERSION}/${CNI_TAR}"
190+
191+
echo "Downloading CNI plugins from: $CNI_URL"
192+
193+
# Download CNI tarball and checksum
194+
curl -fsSL -o ${CNI_TAR} ${CNI_URL}
195+
196+
sudo mkdir -p /opt/cni/bin
197+
sudo tar Cxzvf /opt/cni/bin ${CNI_TAR}
198+
199+
# Configure containerd
200+
sudo mkdir -p /etc/containerd
201+
202+
# Generate base configuration
203+
sudo containerd config default | sudo tee /etc/containerd/config.toml > /dev/null
204+
205+
# Configure based on version
206+
if [ "$MAJOR_VERSION" = "2" ]; then
207+
# Containerd 2.x configuration
208+
cat <<EOF | sudo tee /etc/containerd/config.toml > /dev/null
209+
version = 2
210+
root = "/var/lib/containerd"
211+
state = "/run/containerd"
212+
213+
[grpc]
214+
address = "/run/containerd/containerd.sock"
215+
uid = 0
216+
gid = 0
217+
218+
[plugins]
219+
[plugins."io.containerd.grpc.v1.cri"]
220+
sandbox_image = "registry.k8s.io/pause:3.9"
221+
systemd_cgroup = true
222+
[plugins."io.containerd.grpc.v1.cri".containerd]
223+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes]
224+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
225+
runtime_type = "io.containerd.runc.v2"
226+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
227+
SystemdCgroup = true
228+
[plugins."io.containerd.grpc.v1.cri".registry]
229+
[plugins."io.containerd.grpc.v1.cri".registry.mirrors]
230+
[plugins."io.containerd.grpc.v1.cri".registry.mirrors."docker.io"]
231+
endpoint = ["https://registry-1.docker.io"]
232+
EOF
233+
else
234+
# Containerd 1.x configuration
235+
cat <<EOF | sudo tee /etc/containerd/config.toml > /dev/null
236+
version = 1
237+
root = "/var/lib/containerd"
238+
state = "/run/containerd"
239+
240+
[grpc]
241+
address = "/run/containerd/containerd.sock"
242+
uid = 0
243+
gid = 0
244+
245+
[plugins]
246+
[plugins."io.containerd.grpc.v1.cri"]
247+
sandbox_image = "registry.k8s.io/pause:3.9"
248+
systemd_cgroup = true
249+
[plugins."io.containerd.grpc.v1.cri".containerd]
250+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes]
251+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
252+
runtime_type = "io.containerd.runtime.v1.linux"
253+
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
254+
SystemdCgroup = true
255+
[plugins."io.containerd.grpc.v1.cri".registry]
256+
[plugins."io.containerd.grpc.v1.cri".registry.mirrors]
257+
[plugins."io.containerd.grpc.v1.cri".registry.mirrors."docker.io"]
258+
endpoint = ["https://registry-1.docker.io"]
259+
EOF
260+
fi
261+
262+
# Set up systemd service for containerd
263+
sudo curl -fsSL "https://raw.githubusercontent.com/containerd/containerd/main/containerd.service" -o /etc/systemd/system/containerd.service
264+
265+
# Create containerd service directory
266+
sudo mkdir -p /etc/systemd/system/containerd.service.d
267+
268+
# Add custom service configuration
269+
cat <<EOF | sudo tee /etc/systemd/system/containerd.service.d/override.conf > /dev/null
270+
[Service]
271+
ExecStart=
272+
ExecStart=/usr/local/bin/containerd
273+
Restart=always
274+
RestartSec=5
275+
Delegate=yes
276+
KillMode=process
277+
OOMScoreAdjust=-999
278+
LimitNOFILE=1048576
279+
LimitNPROC=infinity
280+
LimitCORE=infinity
281+
TasksMax=infinity
282+
# Ensure socket directory exists with correct permissions
283+
ExecStartPre=/bin/mkdir -p /run/containerd
284+
ExecStartPre=/bin/chmod 711 /run/containerd
285+
EOF
286+
287+
# Reload systemd and start containerd
288+
sudo systemctl daemon-reload
289+
sudo systemctl enable --now containerd
290+
291+
# Wait for containerd to be ready
292+
echo "Waiting for containerd to be ready..."
293+
timeout=60
294+
while ! sudo ctr version &>/dev/null; do
295+
if [ $timeout -le 0 ]; then
296+
echo "Timeout waiting for containerd to be ready"
297+
exit 1
298+
fi
299+
sleep 1
300+
timeout=$((timeout-1))
301+
done
302+
303+
# Ensure socket permissions are correct
304+
sudo chmod 666 /run/containerd/containerd.sock
305+
306+
# Verify installation
307+
echo "Verifying installation..."
308+
containerd --version
309+
runc --version
310+
sudo ctr version
311+
312+
# Test containerd functionality
313+
echo "Testing containerd functionality..."
314+
sudo ctr images pull docker.io/library/hello-world:latest
315+
sudo ctr run --rm docker.io/library/hello-world:latest test
316+
317+
echo "Containerd installation completed successfully!"
58318
`
59319

60320
type Containerd struct {

0 commit comments

Comments
 (0)