3.1 Control Plane vs Agent Execution Plane
3.2 Project Scope Model
3.3 Data Flow: Source -> Release -> Pipeline -> Logs/Artifacts
3.4 Runtime Configuration Layers (global / project / environment)
3.5 Pipeline Execution Semantics
3.6 Release Governance Path
3.7 Rollback Architecture (Policy-driven)
3.8 Security and Trust Boundaries
3.9 State and Persistence Model
3.10 Scalability Model
3.11 Failure Modes and Recovery Patterns
3.12 Why This Architecture Works in Practice
21.1 Agent Not Claiming Jobs
Orbnetes deployment and release orchestration documentation for operators and platform teams.
This section covers high-frequency operational issues and practical resolution steps. Use it as a fast diagnostic reference before deep incident escalation.
Symptoms
- Jobs stay in
queued. - Agent appears online, but execution never starts.
- Queue grows while expected runner seems available.
Most common causes
- Job tags do not match agent tags.
- Agent is online but not allowed for current project.
- Agent is inactive/disabled for scheduling.
- Dependencies (
needs) are blocking candidate job. - Approval gate is pending (release not yet executable).
Quick checks
- Verify job tags in blueprint/job metadata.
- Verify agent tags on agent page.
- Verify project Allowed Agents includes that agent.
- Verify agent status is online (fresh heartbeat).
- Verify release is not in
pending_approval. - Verify upstream dependencies succeeded.
Fix actions
- Align tags and relaunch/rerun.
- Add agent to project allowed set.
- Reactivate agent if inactive.
- Resolve approval/dependency blockers.