autonomy ha split-brain detect¶
Detect split-brain conditions (epoch divergence, unclosed epochs)
Synopsis¶
Queries the control-plane split-brain detection endpoint and displays:
Overall risk level: none / possible / detected
Epoch divergence between this node’s in-process state and the durable DB epoch
Unclosed epoch count (expected: 0 or 1; > 1 indicates an anomaly)
Holder ID consistency between the local elector and the DB record
Recommended next action when risk is non-zero
Split-brain in this architecture means a control-plane node believes it has write authority (holds the advisory lock and cached epoch) but the durable epoch in PostgreSQL has been incremented by a new leader. Writes by the stale node will be rejected by EpochFence, but the node may be confused about its own state.
Risk levels: none — no indicators; cluster is operating normally possible — anomaly detected (e.g. unclosed epochs); investigate recommended detected — clear epoch divergence; stale leader present; action required
Recommended actions:
none — no action needed
investigate — review epoch history: autonomy ha status
force_failover — trigger failover from a healthy node:
autonomy ha failover trigger –operator
This command is read-only and safe to run from any node. It does not enforce write-blocking or trigger recovery.
Usage¶
autonomy ha split-brain detect [flags]
Options¶
--orchestrator-url string orchestrator base URL (env: AUTONOMY_ORCHESTRATOR_URL)
--output string output format: "text" (human-readable) or "json" (machine-readable) (default "text")
See also¶
autonomy ha split-brain— Split-brain detection and recovery operations