Security Configuration Suggestions for Using Containers in CCE Clusters
Controlling the Pod Scheduling Scope
The nodeSelector or nodeAffinity is used to limit the range of nodes to which applications can be scheduled, preventing the entire cluster from being threatened due to the exceptions of a single application. For details, see Node Affinity.
Suggestions on Container Security Configuration
- Set the computing resource limits (request and limit) of a container. This prevents the container from occupying too many resources and affecting the stability of the host and other containers on the same node.
- Unless necessary, do not mount sensitive host directories to containers, such as /, /boot, /dev, /etc, /lib, /proc, /sys, and /usr.
- Do not run the sshd process in containers unless necessary.
- Unless necessary, it is not recommended that containers and hosts share the network namespace.
- Unless necessary, it is not recommended that containers and hosts share the process namespace.
- Unless necessary, it is not recommended that containers and hosts share the IPC namespace.
- Unless necessary, it is not recommended that containers and hosts share the UTS namespace.
- Unless necessary, do not mount the sock file of Docker to any container.
Container Permission Access Control
When using a containerized application, comply with the minimum privilege principle and properly set securityContext of Deployments or StatefulSets.
- Configure runAsUser to specify a non-root user to run a container.
- Configure privileged to prevent containers being used in scenarios where privilege is not required.
- Configure capabilities to accurately control the privileged access permission of containers.
- Configure allowPrivilegeEscalation to disable privilege escape in scenarios where privilege escalation is not required for container processes.
- Configure seccomp to restrict the container syscalls. For details, see Restrict a Container's Syscalls with seccomp in the official Kubernetes documentation.
- Configure ReadOnlyRootFilesystem to protect the root file system of a container.
Example YAML for a Deployment:
apiVersion: apps/v1 kind: Deployment metadata: name: security-context-example namespace: security-example spec: replicas: 1 selector: matchLabels: app: security-context-example label: security-context-example strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: annotations: seccomp.security.alpha.kubernetes.io/pod: runtime/default labels: app: security-context-example label: security-context-example spec: containers: - image: ... imagePullPolicy: Always name: security-context-example securityContext: allowPrivilegeEscalation: false readOnlyRootFilesystem: true runAsUser: 1000 capabilities: add: - NET_BIND_SERVICE drop: - all volumeMounts: - mountPath: /etc/localtime name: localtime readOnly: true - mountPath: /opt/write-file-dir name: tmpfs-example-001 securityContext: seccompProfile: type: RuntimeDefault volumes: - hostPath: path: /etc/localtime type: "" name: localtime - emptyDir: {} name: tmpfs-example-001
Restricting the Access of Containers to the Management Plane
If application containers on a node do not need to access Kubernetes, you can perform the following operations to disable containers from accessing kube-apiserver:
- Query the container CIDR block and private API server address.
On the Clusters page of the CCE console, click the name of the cluster to find the information on the details page.
- Configure access rules.
- CCE cluster: Log in to each node in the cluster as user root and run the following command:
- VPC network:
iptables -I OUTPUT -s {container_cidr} -d {Private API server IP} -j REJECT
- Container tunnel network:
iptables -I FORWARD -s {container_cidr} -d {Private API server IP} -j REJECT
{container_cidr} indicates the container CIDR of the cluster, for example, 10.0.0.0/16.
To ensure configuration persistence, you are advised to write the command to the /etc/rc.local script.
- VPC network:
- CCE Turbo cluster: Add an outbound rule to the ENI security group of the cluster.
- Log in to the VPC console.
- In the navigation pane, choose Access Control > Security Groups.
- Locate the ENI security group corresponding to the cluster and name it in the format of {Cluster name}-cce-eni-{Random ID}. Click the security group name and configure rules.
- Click the Outbound Rules tab and click Add Rule to add an outbound rule for the security group.
- Priority: Set it to 1.
- Action: Select Deny, indicating that the access to the destination address is denied.
- Type: Select IPv4.
- Protocol & Port: Enter 5443 based on the port in the intranet API server address.
- Destination: Select IP address and enter the IP address of the internal API server.
- Click OK.
- CCE cluster: Log in to each node in the cluster as user root and run the following command:
- Run the following command in the container to access kube-apiserver and check whether the request is intercepted:
curl -k https://{Private API server IP}:5443
Feedback
Was this page helpful?
Provide feedbackThank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot