How clusterlabs/pacemaker works
Pacemaker is a high-availability cluster resource manager. Administrators configure resources and constraints; Pacemaker coordinates cluster membership and configuration, computes a desired resource plan, executes and monitors actions, and recovers from failures, including fencing nodes when needed. The graph preserves the major daemon and library responsibilities visible in the tree. Many implementation excerpts were not provided, so most detailed runtime links are marked dashed and unverified rather than presented as inspected call paths.
Main entry points: based Configuration Service · controld Cluster Controller · pcmk_scheduler.c Action Scheduler · controld_transition.c Transition Engine
Components 18
Administration
- Cluster Tools
tools - Cluster Monitor
tools/crm_mon.c
Cluster Coordination
- Configuration Service
daemons/based - Cluster Controller
daemons/controld - Membership and Messaging
lib/cluster - Attribute Service
daemons/attrd
Planning and State
- Cluster Configuration
daemons/based - CIB Client API
lib/cib/cib_client.c - Action Scheduler
lib/pacemaker/pcmk_scheduler.c - Status Unpacking
lib/pengine/unpack.c - Transition Engine
daemons/controld/controld_transition.c
Execution and Recovery
- Resource Executor
daemons/execd/execd_commands.c - Agent Execution
lib/services/services.c - Platform Backends
lib/services/services_linux.c - Fencing Service
daemons/fenced/fenced_commands.c - Fencing Client API
lib/fencing/st_client.c - Alerting
daemons/execd/execd_alerts.c
Other components
- Administrator
Connections 18
- Administrator → Cluster Tools: configures
- Cluster Tools → CIB Client API: submits requests
- CIB Client API → Configuration Service: connects to
- Configuration Service → Cluster Configuration: reads and updates
- Cluster Controller → Membership and Messaging: tracks membership
- Cluster Controller → Configuration Service: uses configuration
- Cluster Controller → Attribute Service: shares attributes
- Configuration Service → Action Scheduler: requests planning
- Action Scheduler → Status Unpacking: interprets status
- Action Scheduler → Transition Engine: provides actions
- Transition Engine → Resource Executor: dispatches actions
- Resource Executor → Agent Execution: runs agents
- Agent Execution → Platform Backends: selects backend
- Transition Engine → Fencing Client API: requests fencing
- Fencing Client API → Fencing Service: sends commands
- Resource Executor → Alerting: handles alerts
- Cluster Monitor → CIB Client API: queries cluster
- Cluster Monitor → Administrator: reports status