ros2_medkit_log_bridge

Drop-in bridge that promotes ROS 2 /rosout log entries to structured medkit faults, attributing each fault to the node that logged it. No changes to the user’s nodes are required.

It is a compatibility adapter, the same category as ros2_medkit_diagnostic_bridge. Native ros2_medkit_fault_reporter instrumentation stays the canonical path for code you control; this bridge is the fallback for nodes that only log.

What it does

Subscribes to /rosout (rcl_interfaces/msg/Log) and forwards entries at or above a severity floor to the FaultManager:

• source_id of each fault is the originating node’s fully-qualified name. It is derived from Log.name by taking the first dotted segment (a Log.name may carry a sub-logger suffix, e.g. controller_manager.resource_manager, and node names cannot contain .) and prefixing /, giving e.g. /controller_manager. The gateway discovers entities by node FQN, so this is the form that lets a fault (and its snapshots / rosbag) associate with the entity in the SOVD tree. Each node gets its own per-node FaultReporter and therefore its own client-side debounce.
• fault_code is auto-generated as <PREFIX>_<NODE>_<HASH>. <HASH> is a fixed FNV-1a 32-bit digest (8 lowercase hex) of a normalized message template (numbers / hex / paths stripped, isolated single-letter tokens dropped) so the same logical message maps to the same code across occurrences. <NODE> is the upper-snake of source_id. The 8-hex hash is never truncated; if the 64-char cap is hit the node part is trimmed instead.

Namespaced-node limitation: Log.name encodes a node’s namespace with the same . separator as a sub-logger suffix, so the two are indistinguishable from the string alone. source_id takes the first dotted segment, which is right for a non-namespaced node with a sub-logger but collapses a namespaced node (robot1.planner_server -> /robot1) to its namespace, so same-named nodes in different namespaces share one code. Multi-robot fleets typically isolate robots by ROS_DOMAIN_ID (one gateway per robot, federated by peer aggregation), which sidesteps this.

Forwarding, the LocalFilter, and confirmation

Two independent debounces sit between a log line and a confirmed fault:

1. Per-node FaultReporter LocalFilter (client-side). WARN is held until default_threshold (3) occurrences within default_window_sec (10s). ERROR/FATAL have severity >= bypass_severity (2) and bypass the filter, forwarding immediately.
2. Bridge report_cooldown_sec cooldown, applied only to ERROR/FATAL (the levels that bypass the LocalFilter). It forwards the first occurrence of a (fault_code, severity) immediately and suppresses that same pair for report_cooldown_sec (default 5s, 0.0 disables), bounding a flood. WARN is never cooled here (that would starve its LocalFilter threshold counting), and keying on severity means a WARN never suppresses a same-message ERROR escalation.

Whether a forwarded fault then shows as PREFAILED (suspected) or CONFIRMED is a separate, gateway-side decision driven by the FaultManager’s confirmation_threshold - not by this bridge and not by the client-side LocalFilter. For visible-but-quiet WARNs, launch the FaultManager with a low confirmation_threshold (or an entity threshold for LOG_* codes).

Hard limitations (by construction)

• Only sees logs that reach /rosout via rclcpp from a still-alive node. Console-only loggers (e.g. some Micro XRCE-DDS / non-rclcpp loggers) are invisible.
• A node that crashes hard may not flush its final log to /rosout, so the terminating ERROR can be missed. Process-death detection belongs to a separate liveliness bridge, not here.

Run it

# next to an existing stack + the medkit gateway/fault_manager
ros2 launch ros2_medkit_log_bridge log_bridge.launch.py

Configuration (config/log_bridge.yaml)

exclude_nodes / include_only_nodes match as unanchored substrings against the node FQN: planner matches /planner_server and /robot1/planner_server. Use a longer, more specific substring (e.g.

File truncated at 100 lines see the full file

Changelog for package ros2_medkit_log_bridge

Forthcoming

• Initial release: promote /rosout log entries (WARN/ERROR/FATAL) to FaultManager faults, attributed to the originating node via a per-source FaultReporter, with auto-generated stable fault codes.