Kubernetes

Check out Jibril's public recipes repository at https://github.com/garnet-org/jibril-balag.

Deploy Jibril on Kubernetes Clusters

To deploy Jibril as a DaemonSet on Kubernetes clusters, use the setup-k8s.sh script (also available at GitHub. This script automatically creates a Deployment file with the necessary ConfigMap, DaemonSet, and related resources.

Make sure to use --dry-run so it does not apply the deployment automatically.

Currently almost all development-like Kubernetes distributions (Minikube, Microk8s, ...) are supported, as long as compute nodes are virtual-machines or real hosts.

Container based compute nodes distributions, like Kind, will make resource consumption bigger and is unsupported for now).

Usage

$ ./setup-k8s.sh [OPTIONS]

Options

--namespace=NAME	Kubernetes namespace Default: security
--image=IMAGE	Jibril container image Default: garnetlabs/jibril:v1.4
--log-level=LEVEL	Log level (quiet, fatal, error, warn, info, debug) Default: info
--config=FILE	Path to custom Jibril config.yaml file Defaullt: built-in
--memory-request=SIZE	Memory request Default: 256Mi
--memory-limit=SIZE	Memory limit Default: 512Mi
--cpu-request=AMOUNT	CPU request Default: 100m
--cpu-limit=AMOUNT	CPU limit Default: 500m
--node-selector=EXPR	Node selector expression (e.g. 'role=security')
--toleration=KEY:VAL:EFFECT	Add toleration (can be used multiple times)
--output=FILE	Output YAML to file Default: jibril-k8s.yaml
--dry-run	Print configuration without applying
--cleanup	Remove existing Jibril resources from the cluster
--help	Show help

Examples

1. Basic deployment with defaults

   $ ./setup-k8s.sh

2. Deploy to a custom namespace

   $ ./setup-k8s.sh --namespace=monitoring

3. Add node toleration

   $ ./setup-k8s.sh --toleration=security-agent:true:NoSchedule

4. Set custom memory limits

   $ ./setup-k8s.sh --memory-limit=1Gi --memory-request=512Mi

5. Target specific nodes with a node selector

   $ ./setup-k8s.sh --node-selector=role=security

6. Deploy on GPU nodes with higher CPU limits

   $ ./setup-k8s.sh --node-selector=gpu=true --cpu-limit=2 --cpu-request=500m

7. Configure multiple tolerations

   $ ./setup-k8s.sh --toleration=security:true:NoSchedule --toleration=critical:true:NoExecute

8. Use a custom Jibril configuration file

   $ ./setup-k8s.sh --config=/path/to/my-jibril-config.yaml

9. Preview configuration without applying

   $ ./setup-k8s.sh --dry-run

10. Save configuration to a custom file

    $ ./setup-k8s.sh --output=jibril-prod.yaml

11. Clean up existing deployment

    $ ./setup-k8s.sh --cleanup --namespace=security

12. Complete production deployment example

    $ ./setup-k8s.sh --namespace=security-prod \
      --image=garnetlabs/jibril:latest \
      --config=/etc/jibril/prod-config.yaml \
      --memory-limit=2Gi \
      --memory-request=1Gi \
      --cpu-limit=1 \
      --toleration=security-monitoring:true:NoSchedule \
      --node-selector=security-tier=high

Notes

• Jibril requires privileged access to run eBPF programs.

• The script mounts necessary paths from the host:

  • /sys/fs/bpf

  • /sys/kernel/debug

  • /sys

  • /proc

  • /var/log/jibril

• Log files are stored in /var/log/jibril on the host.

• Configuration is supplied via a ConfigMap.