wehub/yuan-lab-llm--clawmanager
1
0
Fork 0
Code Issues Pull Requests Actions 1 Deployments Models Agent Notes Packages Projects Releases Wiki Activity
Files
main
yuan-lab-llm--clawmanager/docs/deployment.md
T
wehub-resource-sync 6d978fe483
Build / Frontend Build (push) Has been cancelled
Details
Build / Backend Build (push) Has been cancelled
Details
Build / Docker Build (push) Has been cancelled
Details
Build / Kubernetes Deploy Smoke Test (push) Has been cancelled
Details
chore: import upstream snapshot with attribution
2026-07-13 12:35:24 +08:00

5.8 KiB
Raw Permalink Blame History

Deployment Guide

ClawManager is packaged as a Kubernetes-first platform. This guide is the operational entry point for deploying the control plane, locating the relevant manifests in the repository, and understanding which services are expected to come up in a working environment.

Deployment Paths

Choose the deployment path that matches your environment:

The cluster profile is validated with Longhorn as the example CSI implementation. It uses longhorn for RWO control-plane and instance volumes, and longhorn-rwx for the RWX workspace volume. These names are not a project requirement; replace them with compatible StorageClasses for your storage provider.

What Gets Deployed

  • ClawManager frontend and backend
  • MySQL for application state
  • MinIO for object storage-backed features
  • skill-scanner for skill analysis workflows
  • Kubernetes Services used for portal, gateway, and supporting traffic paths

Repository Entry Points

Deployment Workflow

  1. Choose the Kubernetes distribution: k3s or k8s.
  2. Choose the storage profile: single-node or cluster.
  3. Check the storage prerequisites for that profile.
  4. Review the bundled manifest and adjust secrets, images, StorageClass names, and ingress exposure for your environment.
  5. Deploy the platform components into the cluster.
  6. Wait for the core services to become ready.
  7. Validate frontend access, AI Gateway management pages, Security Center connectivity, and runtime creation flows.

Single-node example:

kubectl get nodes
kubectl label node <node> clawmanager.io/storage-node=true --overwrite
kubectl apply -f deployments/k8s/single-node/clawmanager.yaml
kubectl get pvc -n clawmanager-system
kubectl get pods -n clawmanager-system

Cluster example:

kubectl get storageclass longhorn longhorn-rwx
kubectl apply -f deployments/k8s/cluster/clawmanager.yaml
kubectl get pvc -n clawmanager-system
kubectl get pods -n clawmanager-system

Storage Profiles

Single-Node

The single-node profile is the official HostPath validation path. Label exactly one node with clawmanager.io/storage-node=true before installation. The manifest pins HostPath PVs through node affinity and runs clawmanager-app as a single replica. MySQL, Redis, MinIO, workspace, and runtime data are all backed by persistent volumes; durable data must not use emptyDir.

Cluster

The cluster profile is the official multi-node CSI/RWX validation path. The bundled manifest uses Longhorn as an example only:

  • longhorn: RWO MySQL, Redis, MinIO, and instance volumes
  • longhorn-rwx: RWX workspace volume shared by ClawManager and runtime Pods

Set these environment variables in the ClawManager app deployment when replacing the storage provider:

  • CLAWMANAGER_STORAGE_PROFILE=cluster
  • K8S_HOSTPATH_FALLBACK_ENABLED=false
  • K8S_PVC_BIND_TIMEOUT=2m
  • K8S_CONTROL_PLANE_STORAGE_CLASS=<rwo-storage-class>
  • K8S_INSTANCE_STORAGE_CLASS=<rwo-storage-class>
  • K8S_WORKSPACE_STORAGE_CLASS=<rwx-storage-class>
  • K8S_WORKSPACE_ACCESS_MODE=ReadWriteMany

Unsupported combinations:

  • multi-node HostPath as a production or shared workspace strategy
  • local-path or other node-local storage pretending to provide RWX across nodes
  • cluster-internal Service DNS such as workspace-store.clawmanager-system.svc.cluster.local as an NFS server for kubelet-mounted PVs
  • durable MySQL, Redis, MinIO, workspace, or object data on emptyDir
  • cluster profile with implicit HostPath fallback

Operational Notes

  • ClawManager is designed around in-cluster services and platform-mediated access rather than direct pod exposure.
  • Resource Management features depend on object storage and skill-scanner being available.
  • Runtime workspace .openclaw and .hermes archive import/export size is controlled by CLAWMANAGER_WORKSPACE_ARCHIVE_MAX_MIB. The default is 500 MiB; set the env var on the ClawManager app deployment when a larger or smaller limit is needed.
  • For install issues, collect kubectl get storageclass, kubectl get pvc -n clawmanager-system, kubectl get pods -n clawmanager-system, kubectl get events -n clawmanager-system --sort-by=.lastTimestamp, and kubectl describe pvc -n clawmanager-system <pvc-name> output before filing an issue.
  • Production environments should review images, credentials, TLS, persistence, and networking policies before rollout.

Related Guides

Reference in New Issue View Git Blame Copy Permalink