3. Configuration¶

DDS Recorder Configuration
Fast DDS Configuration

3.1. DDS Recorder Configuration¶

A DDS Recorder is configured by a .yaml configuration file. This .yaml file contains all the information regarding the DDS interface configuration, recording parameters, and DDS Recorder specifications. Thus, this file has four major configuration groups:

dds: configuration related to DDS communication.
recorder: configuration of data writing in the database.
remote-controller: configuration of the remote controller of the DDS Recorder.
specs: configuration of the internal operation of the DDS Recorder.

3.1.1. DDS Configuration¶

Configuration related to DDS communication.

3.1.1.1. DDS Domain¶

Tag domain configures the Domain Id.

domain: 101

3.1.1.2. Built-in Topics¶

The discovery phase can be accelerated by listing topics under the builtin-topics tag. The DDS Recorder will create the DataWriters and DataReaders for these topics in the DDS Recorder initialization. The Topic QoS for these topics can be manually configured with the Manual Topic and with the Specs Topic QoS; if a Topic QoS is not configured, it will take its default value.

The builtin-topics must specify a name and type without wildcard characters.

Example of usage:

builtin-topics:
  - name: HelloWorldTopic
    type: HelloWorld

3.1.1.3. Topic Filtering¶

The DDS Recorder automatically detects the topics that are being used in a DDS Network. The DDS Recorder then creates internal DDS Readers to record the data published on each topic. The DDS Recorder allows filtering DDS Topics to allow users to configure the DDS Topics that must be recorded. These data filtering rules can be configured under the allowlist and blocklist tags. If the allowlist and blocklist are not configured, the DDS Recorder will recorded the data published on every topic it discovers. If both the allowlist and blocklist are configured and a topic appears in both of them, the blocklist has priority and the topic will be blocked.

Topics are determined by the tags name (required) and type, both of which accept wildcard characters.

Note

Placing quotation marks around values in a YAML file is generally optional, but values containing wildcard characters do require single or double quotation marks.

Consider the following example:

allowlist:
  - name: AllowedTopic1
    type: Allowed

  - name: AllowedTopic2
    type: "*"

  - name: HelloWorldTopic
    type: HelloWorld

blocklist:
  - name: "*"
    type: HelloWorld

In this example, the data published in the topic AllowedTopic1 with type Allowed and in the topic AllowedTopic2 with any type will be recorded by the DDS Recorder. The data published in the topic HelloWorldTopic with type HelloWorld will be blocked, since the blocklist is blocking all topics with any name and with type HelloWorld.

3.1.1.4. Topic QoS¶

The following is the set of QoS that are configurable for a topic. For more information on topics, please read the Fast DDS Topic section.

Quality of Service	Yaml tag	Data type	Default value	QoS set
Reliability	`reliability`	bool	`false`	`RELIABLE` / `BEST_EFFORT`
Durability	`durability`	bool	`false`	`TRANSIENT_LOCAL` / `VOLATILE`
Ownership	`ownership`	bool	`false`	`EXCLUSIVE_OWNERSHIP_QOS` / `SHARED_OWNERSHIP_QOS`
Partitions	`partitions`	bool	`false`	Topic with / without partitions
Key	`keyed`	bool	`false`	Topic with / without key
History Depth	`history-depth`	unsigned integer	`5000`	History Depth
Max Reception Rate	`max-rx-rate`	float	`0` (unlimited)	Max Reception Rate
Downsampling	`downsampling`	unsigned integer	`1`	Downsampling

Warning

Manually configuring TRANSIENT_LOCAL durability may lead to incompatibility issues when the discovered reliability is BEST_EFFORT. Please ensure to always configure the reliability when configuring the durability to avoid the issue.

3.1.1.4.1. History Depth¶

The history-depth tag configures the history depth of the Fast DDS internal entities. By default, the depth of every RTPS History instance is 5000, which sets a constraint on the maximum number of samples a DDS Recorder instance can deliver to late joiner Readers configured with TRANSIENT_LOCAL DurabilityQosPolicyKind. Its value should be decreased when the sample size and/or number of created endpoints (increasing with the number of topics) are big enough to cause memory exhaustion issues. If enough memory is available, however, the history-depth could be increased to deliver a greater number of samples to late joiners.

3.1.1.4.2. Max Reception Rate¶

The max-rx-rate tag limits the frequency [Hz] at which samples are processed by discarding messages received before 1/max-rx-rate seconds have passed since the last processed message. It only accepts non-negative numbers. By default it is set to 0; it processes samples at an unlimited reception rate.

3.1.1.4.3. Downsampling¶

The downsampling tag reduces the sampling rate of the received data by only keeping 1 out of every n samples received (per topic), where n is the value specified under the downsampling tag. When the max-rx-rate tag is also set, downsampling only applies to messages that have passed the max-rx-rate filter. It only accepts positive integers. By default it is set to 1; it accepts every message.

3.1.1.5. Manual Topics¶

A subset of Topic QoS can be manually configured for a specific topic under the tag topics. The tag topics has a required name tag that accepts wildcard characters. It also has two optional tags: a type tag that accepts wildcard characters, and a qos tag with the Topic QoS that the user wants to manually configure. If a qos is not manually configured, it will get its value by discovery.

topics:
  - name: "temperature/*"
    type: "temperature/types/*"
    qos:
      max-rx-rate: 15
      downsampling: 2

Note

The Topic QoS configured in the Manual Topics take precedence over the Specs Topic QoS.

3.1.1.6. Ignore Participant Flags¶

A set of discovery traffic filters can be defined in order to add an extra level of isolation. This configuration option can be set through the ignore-participant-flags tag:

ignore-participant-flags: no_filter                          # No filter (default)
# or
ignore-participant-flags: filter_different_host              # Discovery traffic from another host is discarded
# or
ignore-participant-flags: filter_different_process           # Discovery traffic from another process on same host is discarded
# or
ignore-participant-flags: filter_same_process                # Discovery traffic from own process is discarded
# or
ignore-participant-flags: filter_different_and_same_process  # Discovery traffic from own host is discarded

See Ignore Participant Flags for more information.

3.1.1.7. Custom Transport Descriptors¶

By default, DDS Recorder internal participants are created with enabled UDP and Shared Memory transport descriptors. The use of one or the other for communication will depend on the specific scenario, and whenever both are viable candidates, the most efficient one (Shared Memory Transport) is automatically selected. However, a user may desire to force the use of one of the two, which can be accomplished via the transport configuration tag.

transport: builtin    # UDP & SHM (default)
# or
transport: udp        # UDP only
# or
transport: shm        # SHM only

Warning

When configured with transport: shm, DDS Recorder will only communicate with applications using Shared Memory Transport exclusively (with disabled UDP transport).

3.1.1.8. Interface Whitelist¶

Optional tag whitelist-interfaces allows to limit the network interfaces used by UDP and TCP transport. This may be useful to only allow communication within the host (note: same can be done with Ignore Participant Flags). Example:

whitelist-interfaces:
  - "127.0.0.1"    # Localhost only

See Interface Whitelist for more information.

3.1.2. Recorder Configuration¶

Configuration of data writing in the database.

3.1.2.1. Output File¶

The recorder output file does support the following configuration settings under the output tag:

Parameter	Tag	Description	Data type	Default value
File path	`path`	Configure the path to save the output file.	`string`	`.`
File name	`filename`	Configure the name of the output file.	`string`	`output`
Timestamp format	`timestamp-format`	Configure the format of the output file timestamp (as in `std::put_time`).	`string`	`%Y-%m-%d_%H-%M-%S_%Z`
Local timestamp	`local-timestamp`	Whether to use a local or global (GMT) timestamp.	`boolean`	`true`
Safety margin	`safety_margin`	Configure safety margin (bytes) used in MCAP file size estimations.	`unsigned int`	`0`
Resource limits	`resource-limits`	Resource Limits	`map`	`unlimited`

When DDS Recorder application is launched (or when remotely controlled, every time a start/pause command is received while in SUSPENDED/STOPPED state), a temporary file with filename name (+timestamp prefix) and .mcap.tmp~ extension is created in path. This file is not readable until the application terminates, receives a suspend/stop/close command, or the file reaches its maximum size (see Resource Limits). On such event, the temporal file is renamed to have .mcap extension in the same location, and is then ready to be processed.

3.1.2.1.1. Resource Limits¶

The resource-limits tag allows users to limit the size of the DDS Recorder’s output. The max-file-size tag specifies the maximum size of each output file and the max-size tag specifies the maximum aggregate size of all output files. If the max-size is higher than the max-file-size, the DDS Recorder will create multiple files with a maximum size of max-file-size. By default, however, the max-file-size is unlimited (0B) and the max-size is the same as the max-file-size; that is, by default the DDS Recorder creates a single file of unlimited size.

Warning

If the max-file-size or the max-size are set to a value lower than the available space in the disk, the DDS Recorder will replace them with the available space in the disk.

To keep the DDS Recorder recording after reaching the max-size, users can set the file-rotation tag to true. Enabling file-rotation allows the DDS Recorder to overwrite old files to free space for new ones.

Note

If the DDS Recorder’s state changes to stop while file-rotation is set, the DDS Recorder will save its current output files from being overwritten in the future.

Note

If an output file is moved, deleted, or renamed, the DDS Recorder will keep the size of the file reserved and rotate between the remaining files.

Example of usage

resource-limits:
  max-file-size: 250KB
  max-size: 2MiB
  file-rotation: true

3.1.2.2. Buffer size¶

buffer-size indicates the number of samples to be stored in the process memory before the dump to disk. This avoids disk access each time a sample is received. By default, its value is set to 100.

3.1.2.3. Event Window¶

DDS Recorder can be configured to continue saving data when it is in paused mode. Thus, when an event is triggered from the remote controller, samples received in the last event-window seconds are stored in the database.

In other words, the event-window acts as a sliding time window that allows to save the collected samples in this time window only when the remote controller event is received. By default, its value is set to 20 seconds.

3.1.2.4. Log Publish Time¶

By default (log-publish-time: false) received messages are stored in the MCAP file with logTime value equals to the reception timestamp. Additionally, the timestamp corresponding to when messages were initially published (publishTime) is also included in the information dumped to MCAP files. In some applications, it may be required to use the publishTime as logTime, which can be achieved by providing the log-publish-time: true configuration option.

3.1.2.5. Only With Type¶

By default, all (allowed) received messages are recorded regardless of whether their associated type information has been received. However, a user can enforce that only samples whose type is received are recorded by setting only-with-type: true.

3.1.2.6. Compression¶

Compression settings for writing to an MCAP file can be specified under the compression configuration tag. The supported compression options are:

Parameter	Tag	Description	Data type	Default value	Possible values
Compression Algorithm	`algorithm`	Compression algorithm to use when writing Chunks.	`string`	`zstd`	`none` `lz4` `zstd`
Compression Level	`level`	Compression level to use when writing Chunks.	`string`	`default`	`fastest` `fast` `default` `slow` `slowest`
Force Compression	`force`	Force compression on all Chunks (even for those that do not benefit from compression).	`boolean`	`false`	`true` `false`

3.1.2.7. Record Types¶

By default, all type information received during execution is stored in a dedicated MCAP file section. This information is then leveraged by DDS Replayer on playback, publishing recorded types in addition to data samples, which may be required for receiver applications relying on Dynamic Types (see Replay Types). However, a user may choose to disable this feature by setting record-types: false.

3.1.2.8. Topic type format¶

The optional ros2-types tag enables specification of the format for storing schemas. When set to true, schemas are stored in ROS 2 message format (.msg). If set to false, schemas are stored in OMG IDL format (.idl). By default it is set to false.

3.1.3. Remote Controller¶

Configuration of the DDS remote control system. Please refer to Remote Control for further information on how to use DDS Recorder remotely. The supported configurations are:

Parameter	Tag	Description	Data type	Default value	Possible values
Enable	`enable`	Enable DDS remote control system.	`boolean`	`true`	`true` `false`
DDS Domain	`domain`	DDS Domain of the DDS remote control system.	`integer`	DDS domain being recorded	From `0` to `255`
Initial state	`initial-state`	Initial state of DDS Recorder.	`string`	`RUNNING`	`RUNNING` `PAUSED` `SUSPENDED` `STOPPED`
Command Topic Name	`command-topic-name`	Name of Controller Command DDS Topic.	`string`	`/ddsrecorder/command`
Status Topic Name	`status-topic-name`	Name of Controller Status DDS Topic.	`string`	`/ddsrecorder/status`

3.1.4. Specs Configuration¶

The internals of a DDS Recorder can be configured using the specs optional tag that contains certain options related with the overall configuration of the DDS Recorder instance to run. The values available to configure are:

3.1.4.1. Number of Threads¶

specs supports a threads optional value that allows the user to set a maximum number of threads for the internal ThreadPool. This ThreadPool allows to limit the number of threads spawned by the application. This improves the performance of the internal data communications.

This value should be set by each user depending on each system characteristics. In case this value is not set, the default number of threads used is 12.

3.1.4.2. Maximum Number of Pending Samples¶

It is possible that a DDS Recorder starts receiving data from a topic that it has not yet registered, i.e. a topic for which it does not know the data type. In this case, messages are kept in an internal circular buffer until their associated type information is received, event on which they are written to disk.

However, the recorder execution might end before this event ever occurs. Depending on configuration (see Only With Type), messages kept in the pending samples buffer will be stored or not on closure. Hence, note that memory consumption would continuously grow whenever a sample with unknown type information is received.

To avoid the exhaustion of memory resources in such scenarios, a configuration option is provided which lets the user set a limit on memory usage. The max-pending-samples parameter allows to configure the size of the aforementioned circular buffers for each topic that is discovered. The default value is equal to 5000 samples, with -1 meaning no limit, and 0 no pending samples.

Depending on the combination of this configuration option and the value of only-with-type, the following situations may arise when a message with unknown type is received:

If max-pending-samples is -1, or if it is greater than 0 and the circular buffer is not full, the sample is added to the collection.
If max-pending-samples is greater than 0 and the circular buffer reaches its maximum capacity, the oldest sample with same type as the received one is popped, and either written without type (only-with-type: false) or discarded (only-with-type: true).
If max-pending-samples is 0, the message is written without type if only-with-type: false, and discarded otherwise.

3.1.4.3. Cleanup Period¶

As explained in Event Window, a DDS Recorder in paused mode awaits for an event command to write in disk all samples received in the last event-window seconds. To accomplish this, received samples are stored in memory until the aforementioned event is triggered and, in order to limit memory consumption, outdated (received more than event-window seconds ago) samples are removed from this buffer every cleanup-period seconds. By default, its value is equal to twice the event-window.

3.1.4.4. QoS¶

specs supports a qos optional tag to configure the default values of the Topic QoS.

Note

The Topic QoS configured in specs can be overwritten by the Manual Topics.

3.1.4.5. Logging¶

specs supports a logging optional tag to configure the DDS Recorder logs. Under the logging tag, users can configure the type of logs to display and filter the logs based on their content and category. When configuring the verbosity to info, all types of logs, including informational messages, warnings, and errors, will be displayed. Conversely, setting it to warning will only show warnings and errors, while choosing error will exclusively display errors. By default, the filter allows all errors to be displayed, while selectively permitting warning and informational messages from DDSRECORDER category.

Note

Configuring the logs via the Command-Line is still active and takes precedence over YAML configuration when both methods are used simultaneously.

Logging	Yaml tag	Description	Data type	Default value	Possible values
Verbosity	`verbosity`	Show messages of equal or higher importance.	enum	`error`	`info` / `warning` / `error`
Filter	`filter`	Regex to filter the category or message of the logs.	string	info : `DDSRECORDER` warning : `DDSRECORDER` error : `""`	Regex string

Note

For the logs to function properly, the -DLOG_INFO=ON compilation flag is required.

The DDS Recorder prints the logs by default (warnings and errors in the standard error and infos in the standard output). The DDS Recorder, however, can also publish the logs in a DDS topic. To publish the logs, under the tag publish, set enable: true and set a domain and a topic-name. The type of the logs published is defined as follows:

LogEntry.idl

const long UNDEFINED = 0x10000000;
const long SAMPLE_LOST = 0x10000001;
const long TOPIC_MISMATCH_TYPE = 0x10000002;
const long TOPIC_MISMATCH_QOS = 0x10000003;
const long FAIL_MCAP_CREATION = 0x12000001;
const long FAIL_MCAP_WRITE = 0x12000002;

enum Kind {
  Info,
  Warning,
  Error
};

struct LogEntry {
  @key long event;
  Kind kind;
  string category;
  string message;
  string timestamp;
};

Note

The type of the logs can be published by setting publish-type: true.

Example of usage

logging:
  verbosity: info
  filter:
    error: "DDSPIPE|DDSRECORDER"
    warning: "DDSPIPE|DDSRECORDER"
    info: "DDSRECORDER"
  publish:
    enable: true
    domain: 84
    topic-name: "DdsRecorderLogs"
    publish-type: false
  stdout: true

3.1.4.6. Monitor¶

specs supports a monitor optional tag to publish internal data from the DDS Recorder. If the monitor is enabled, it publishes (and logs under the MONITOR_DATA log filter) the DDS Recorder’s internal data on a domain, under a topic-name, once every period (in milliseconds). If the monitor is not enabled, the DDS Recorder will not collect or publish any data.

Note

The data published is relative to each period. The DDS Recorder will reset its tracked data after publishing it.

In particular, the DDS Recorder can monitor its internal status and its topics. When monitoring its internal status, the DDS Recorder will track different errors of the DDS Recorder. The type of the data published is defined as follows:

DdsRecorderMonitoringStatus.idl

struct MonitoringErrorStatus {
    boolean type_mismatch;
    boolean qos_mismatch;
};

struct MonitoringStatus {
    MonitoringErrorStatus error_status;
    boolean has_errors;
};

struct DdsRecorderMonitoringErrorStatus {
    boolean mcap_file_creation_failure;
    boolean disk_full;
};

struct DdsRecorderMonitoringStatus : MonitoringStatus {
    DdsRecorderMonitoringErrorStatus ddsrecorder_error_status;
};

When monitoring its topics, the DDS Recorder will track the number of messages lost, received, and the message reception rate [Hz] of each topic. It will also track if a topic’s type is discovered, if there is a type mismatch, and if there is a QoS mismatch. The type of the data published is defined as follows:

MonitoringTopics.idl

struct DdsTopicData
{
    string participant_id;
    unsigned long msgs_lost;
    unsigned long msgs_received;
    double msg_rx_rate;
};

struct DdsTopic
{
    string name;
    string type_name;
    boolean type_discovered;
    boolean type_mismatch;
    boolean qos_mismatch;
    sequence<DdsTopicData> data;
};

struct MonitoringTopics
{
    sequence<DdsTopic> topics;
};

Example of usage

monitor:
  domain: 10
  status:
    enable: true
    domain: 11
    period: 2000
    topic-name: "DdsRecorderStatus"

  topics:
    enable: true
    domain: 12
    period: 1500
    topic-name: "DdsRecorderTopics"

3.1.5. General Example¶

A complete example of all the configurations described on this page can be found below.

Warning

This example can be used as a quick reference, but it may not be correct due to incompatibility or exclusive properties. Do not take it as a working example.

dds:
  domain: 0

  allowlist:
    - name: "topic_name"
      type: "topic_type"

  blocklist:
    - name: "topic_name"
      type: "topic_type"

  builtin-topics:
    - name: "HelloWorldTopic"
      type: "HelloWorld"

  topics:
    - name: "temperature/*"
      type: "temperature/types/*"
      qos:
        max-rx-rate: 15
        downsampling: 2

  ignore-participant-flags: no_filter
  transport: builtin
  whitelist-interfaces:
    - "127.0.0.1"

recorder:
  output:
    filename: "output"
    path: "."
    timestamp-format: "%Y-%m-%d_%H-%M-%S_%Z"
    local-timestamp: false
    safety-margin: 500

    resource-limits:
      max-file-size: 250KB
      max-size: 2MiB
      file-rotation: true

  buffer-size: 50
  event-window: 60
  log-publish-time: false
  only-with-type: false
  compression:
    algorithm: lz4
    level: slowest
    force: true
  record-types: true
  ros2-types: false

remote-controller:
  enable: true
  domain: 10
  initial-state: "PAUSED"
  command-topic-name: "/ddsrecorder/command"
  status-topic-name: "/ddsrecorder/status"

specs:
  threads: 8
  max-pending-samples: 10
  cleanup-period: 90

  qos:
    max-rx-rate: 20
    downsampling: 3

  logging:
    verbosity: info
    filter:
      error: "DDSPIPE|DDSRECORDER"
      warning: "DDSPIPE|DDSRECORDER"
      info: "DDSRECORDER"
    publish:
      enable: true
      domain: 84
      topic-name: "DdsRecorderLogs"
      publish-type: false
    stdout: true

  monitor:
    domain: 10
    topics:
      enable: true
      domain: 11
      period: 1000
      topic-name: "DdsRecorderTopics"

    status:
      enable: true
      domain: 12
      period: 2000
      topic-name: "DdsRecorderStatus"

3.2. Fast DDS Configuration¶

As explained in this section, a DDS Recorder instance stores (by default) all data regardless of whether their associated data type is received or not. Some applications rely on this information being recorded and written in the resulting MCAP file, which requires that the user application is configured to send the necessary type information. However, Fast DDS does not send the data type information by default, it must be configured to do so.

First of all, when generating the topic types using eProsima Fast DDS Gen, the option -typeobject must be added in order to generate the needed code to fill the TypeObject data.

For native types (data types that does not rely in other data types) this is enough, as Fast DDS will send the TypeObject by default. However, for more complex types, it is required to use TypeInformation mechanism. In the Fast DDS DomainParticipant set the following QoS in order to send this information:

DomainParticipantQos pqos;
pqos.wire_protocol().builtin.typelookup_config.use_server = true;

Feel free to review this section, where it is explained in detail how to configure a Fast DDS Publisher/Subscriber leveraging Dynamic Types.