Log Folder Convention

The log in an Isaac application is a directory; not a single file. The directory naming convention is as follows:


The path to the log folder is made up of three parts: the base directory, an application UUID directory and a tag. The base directory needs to be a directory where the user running the application has write privileges. The application UUID is a string representing a unique ID per execution of the application. This is a new unique ID every time the application runs. It cannot be changed or predetermined. Finally, the tag is an optional folder that lets the user distinguish between different logs captured during a single execution of the application.

The only components of the log folder that are user-defined are the base directory and the tag.


The default value for the base directory is /tmp/isaac and in Linux, the contents of /tmp are erased with every reboot.

To save log files to a user-defined folder, the folder must be created first. The Invalid path for log. error indicates that the directory did not exist at the time the log file attempted to save.

Applications that must record a component’s channel must define a node containing a recorder component, and then connect all the components to be recorded to that recorder component.

As in any Isaac SDK application, the graph can be created with C++ or with a JSON configuration file. This section provides procedures for both methods, for record and replay functionality. JSON configuration is recommended however.

A recorder component definition is shown in the following JSON snippet:


{ "name": "recorder", "components": [ { "name": "MessageLedger", "type": "isaac::alice::MessageLedger" }, { "name": "Recorder", "type": "isaac::cask::Recorder" } ] }

To log channels a and b of components X and Y respectively, define the following JSON edges:


{ "source": "<Node Name>/X/a", "target": "recorder/Recorder/a" }, { "source": "<Node Name>/Y/b", "target": "recorder/Recorder/b" }

The initial definition of the recorder component is the following:


"recorder": { "Recorder": { "base_directory": "/tmp/isaac", "tag": "test", "start_recording_automatically": false } }

The fields that can be configured are: the base_directory and the tag, which define the location of the log, and start_recording_automatically, which controls recording at application startup.

To replay a recorded log, the application must define a node containing a replay component and then connect all the destination components to the replay component.

The initial definition of the replay component is the following:


{ "name": "replay", "components": [ { "name": "MessageLedger", "type": "isaac::alice::MessageLedger" }, { "name": "Replay", "type": "isaac::cask::Replay" } ] }

Connect the replay component to all target channels for which the corresponding source channel was recorded. The channels of the components that the replay component is replacing must be disconnected in the JSON also. Continue the example in the previous section for replay, with the following JSON code:


{ "source": "replay/Replay/a", "target": "<Node Name>/M/a" }, { "source": "replay/Replay/a", "target": "<Node Name>/N/a" }, { "source": "replay/Replay/b", "target": "<Node Name>/P/b" }

The channel a sends replayed messages to components M and N, and channel b to component P. Not all recorded channels must be replayed. One channel can be replayed to more than one target.

Same as in the previous section, the Replay node can have its initial configuration:


"replay": { "Replay": { "cask_directory": "/tmp/isaac/e0d7caae-a70a-11e8-8c38-91bfb5eade6f/test000", "replay_time_offset": 0, "use_recorded_message_time": false } }

Set the cask_directory to the log folder. See Log Folder Convention for more information.

The replay_time_offset parameter is an offset in nanoseconds from where to start the replay. Its default value is zero or no offset.

The use_recorded_message_time specifies the time stamp to be used as publish time for a message during the replay. When false it uses the publish time of the replay component, and when true it uses the time stamp from recorded log messages. By default this option is false.

To replay a log multiple times during a single application execution without killing and restarting, the use_recorded_message_time parameter should be set to false to avoid unexpected behavior during replay. In order to disable automatic start of the replay component, omit cask_directory from the configuration.

© Copyright 2018-2020, NVIDIA Corporation. Last updated on Feb 1, 2023.