Class LogProvider

• Defined in File log_provider.hpp

Class Documentation

class LogProvider

Provider interface for log storage backends.

Typed provider interface implemented by plugins alongside GatewayPlugin via multiple inheritance. When a LogProvider plugin is loaded, LogManager delegates all query/config operations to it and stops using the local /rosout ring buffer for serving results.

Multiple plugins implementing LogProvider can be loaded; only the first registered LogProvider is used for queries (same as UpdateProvider). ALL LogProvider plugins receive on_log_entry() calls (observer pattern) - unless the primary provider returns true from manages_ingestion(), in which case the /rosout subscription is never created and on_log_entry() is never called.

Two modes of operation:

• Observer mode (default, manages_ingestion() == false): LogManager subscribes to /rosout and feeds entries to all observers via on_log_entry(). The primary provider replaces the query/config backend; the /rosout ring buffer is still populated unless on_log_entry() returns true.

• Full ingestion (manages_ingestion() == true): The primary provider owns the entire log pipeline - choosing its own source (journald, CloudWatch, custom topics), format, and storage. LogManager skips the /rosout subscription and ring buffer entirely.

See also

GatewayPlugin for the base class all plugins must also implement

See also

LogManager for the subsystem that uses this

Thread safety

All methods may be called from multiple threads concurrently. Implementations must provide their own synchronization.

Example use cases

• OpenTelemetry exporter: implement on_log_entry() to forward to OTLP endpoint

• Database backend: implement get_logs() to query a SQLite/InfluxDB/etc. store

• Log aggregator: combine /rosout with external log sources

• Full ingestion: subscribe to journald/CloudWatch/custom topics directly

Public Functions

virtual ~LogProvider() = default

inline virtual bool manages_ingestion() const

Whether this provider fully manages log ingestion.

When true, LogManager skips the /rosout subscription and ring buffer entirely. The plugin is responsible for sourcing log entries from whatever transport it chooses (journald, CloudWatch, custom ROS 2 topics, etc.).

When false (the default), LogManager subscribes to /rosout and populates the ring buffer as usual; all LogProvider observers receive on_log_entry() calls.

Only the primary LogProvider’s return value is checked (at LogManager construction).

Returns:

true if the plugin owns the entire log pipeline; false for observer mode

virtual std::vector<LogEntry> get_logs(const std::vector<std::string> &node_fqns, bool prefix_match, const std::string &min_severity, const std::string &context_filter, const std::string &entity_id) = 0

Query log entries for a set of node names.

Parameters:

• node_fqns – Node FQNs WITHOUT leading slash (e.g. “powertrain/engine/temp_sensor”)

• prefix_match – If true, treat each entry as a namespace prefix (for Component queries). If false, match exactly (for App queries).

• min_severity – Optional additional severity filter (“debug”,”info”,”warning”,”error”,”fatal”). Empty string = no additional filter beyond entity config.

• context_filter – Optional substring filter applied to the log entry’s node/logger name.

• entity_id – Entity ID for config lookup (determines base severity_filter and max_entries).

Returns:

LogEntry objects sorted by id ascending, capped at config.max_entries. LogManager handles JSON serialization via entry_to_json().

virtual LogConfig get_config(const std::string &entity_id) const = 0

Get the current log configuration for an entity.

Returns:

Entity config (default values if entity was never configured)

virtual std::string update_config(const std::string &entity_id, const std::optional<std::string> &severity_filter, const std::optional<size_t> &max_entries) = 0

Update log configuration for an entity.

Returns:

Empty string on success, error message on validation failure

inline virtual bool on_log_entry(const LogEntry &entry)

Called for each /rosout log entry before it is stored in the default ring buffer.

All LogProvider plugins receive this call (observer pattern), regardless of whether they are the primary query provider.

OR-aggregation: if ANY observer returns true, ring buffer storage is suppressed. All observers are always called regardless of earlier return values.

Example: An OpenTelemetry plugin forwards to OTLP and returns false (keeps ring buffer). A database plugin stores in SQLite, returns true (replaces ring buffer for that entry).

Note

This method is never called when the primary provider’s manages_ingestion() returns true, because the /rosout subscription is not created in that case.

Parameters:

entry – The log entry (name field is WITHOUT leading slash)

Returns:

true to suppress default ring buffer storage; false to allow it (default)