Operations
Troubleshooting
Common failure modes and where to look first.
API Returns 503
Check:
MONGODB_URIis set- MongoDB is reachable from the API process
- database credentials and IP allowlist are valid
- API logs for collection initialization errors
Agent Stuck In Provisioning
Check:
CLOUD_PROVIDERAWS_REGION,EKS_CLUSTER, and the production AWS deployment variables- IAM role permissions
- EKS quota and node availability
AGENT_IMAGE_URL- API logs from
launchCloudInstance - cluster events for the agent namespace
Agent Starts But Never Works
Check:
- daemon logs in the pod
API_URLreachable from the podAGENT_TOKENinjectedRUNNER_URLset for native runtimeAGENTCOMMONS_API_KEYif Agent Commons is required- task exists in MongoDB and is not already running/completed
World Does Not Update
Check:
GET /fleets/:id/world- WebSocket URL from
commonos world stream-url - browser console
- API broadcast logs
- daemon
POST /eventscalls - tenant token used by the WebSocket connection
AXL Messages Do Not Appear
Check:
- daemon registered
axl.peerIdandaxl.multiaddr GET /fleets/:id/peers- local AXL API URL and port in the pod
- target agent has a peer ID
- message route broadcasts to the fleet
CLI Says Not Authenticated
Run:
commonos auth login --key cos_live_xxx --url http://localhost:3001Or set:
export COMMONOS_API_KEY=cos_live_xxx
export COMMONOS_API_URL=http://localhost:3001