Streaming telemetry allows users to monitor network health by efficiently streaming operational data of interest from OcNOS routers. This structured data is transmitted to remote management systems for proactive network monitoring and understanding CPU and memory usage in managed devices for troubleshooting.

A machine learning (ML) database can be created with telemetry data to establish a baseline for normal network operation and predict or mitigate network issues.

In OcNOS various gRPC Network Management Interface (gNMI) Subscription Modes, Telemetry Modes, and Encoding Types are supported, providing efficient network management capabilities.

Streaming Telemetry Dial-In Mode: In this mode, the collector initiates a connection to the target device (OcNOS) and subscribes to receive telemetry data from OcNOS devices.

Streaming Telemetry Dial-Out Mode (Persistent Subscriptions): In this mode, the target (OcNOS) initiates the gRPC tunnel connection to the collector. Once the connection is established, the collector invokes the “Publish" RPC on the target. Subscriptions configured on the target are then streamed on that connection at the specified sample interval. These subscriptions remain active on OcNOS devices as long as the corresponding configuration on the target exists. If the gRPC tunnel connection is interrupted or the target reboots, the gNMI server on the target re-establishes the connection to the gNMI collector, ensuring continued streaming.

Stream Mode: Enables continuous and real-time transmission of telemetry data from OcNOS devices to the monitoring system. The stream mode applies to both the dial-in and dial-out gNMI subscription modes.

Poll Mode: Poll mode subscriptions allow for on-demand data retrieval through a long-lived RPC. Subscribers initiate this mode by sending a Subscribe request message, followed by sending an empty Poll message to receive the desired data.

Once Mode: In Once mode subscription, the OcNOS device responds to a subscribe request with a one-time data retrieval, similar to a get request. Upon receiving the Once mode subscribe request, the device sends back the subscribe response for all subscriptions in the list and terminates the RPC.

gNMI In-Band support in OcNOS enables streaming telemetry data transmission across any one of the default, management, and user-defined VRFs. If no VRF is defined, streaming telemetry is automatically enabled within the default VRF. This enhancement allows network operators to utilize existing data interfaces for efficient in-band telemetry data transmission.

Protocol Buffers (protobuf): Offers a compact binary serialization format for efficient encoding and transmission of structured telemetry data. Protobuf is optimized for performance and bandwidth efficiency.

JavaScript Object Notation (JSON): Provides a human-readable data interchange format commonly used for telemetry data representation. JSON encoding facilitates interoperability and ease of integration with various applications and tools. It adheres to the JSON specification outlined in RFC7159, employing relevant quoting. Consequently, string values are quoted while number values remain unquoted.

JSON-IETF: This variant of JSON encoding adheres to the IETF standards, ensuring consistency and compatibility with industry specifications. JSON_IETF encoded data conforms the rules outlined in RFC7951 for JSON serialization.

OcNOS supports the protobuf, JSON, and JSON-IETF encoding types for both the dial-in and dial-out gNMI subscription modes

Streaming Telemetry Data Models: OcNOS supports IPI native data models and OpenConfig data models, providing standardized representations of network configurations and telemetry data. This support enhances interoperability and facilitates consistent management across heterogeneous network environments.

To limit the impact of telemetry on critical features of the OcNOS target device, certain limits have been implemented for different platform types.

A system is considered high range if it has eight or more CPU cores and is not based on an "Intel Atom" processor. Users can subscribe to a maximum of 100 sensor paths (including Dial-In and Dial-Out subscriptions) at any given time. The minimum supported sample interval is 10 seconds.

A system is considered standard range if it has fewer than eight CPU cores or is based on an "Intel Atom" processor. Users can subscribe to a maximum of 50 sensor paths (including Dial-In and Dial-Out subscriptions) at any given time. The minimum supported sample interval is 90 seconds.

The total count of sensor paths includes the child paths of a subscribe request. For instance, if a subscribe request has four child paths, the total sensor paths count equals five (the given path plus four child paths). Use the show streaming-telemetry command to display the minimum sample interval and the maximum number of sensor paths supported for a platform.

Minimum Sample Interval: The minimum supported sample interval is 10 seconds for “High-end” platform and 90 seconds for “Standard” platform type. Any sampling mode request with a sample interval of less than the minimum allowed will result in an error. However, if a sample interval is 0, it defaults to the minimum sample interval supported by the gNMI server for that platform type.

gNMI Specification can be found at: https://github.com/openconfig/reference/blob/master/rpc/gnmi/gnmi-specification.md.

This section lists the telemetry commands.

Use this command to enable or disable debugging information for CML streaming telemetry.

None

By default, debugging information is disabled.

Exec Mode

This command was introduced in OcNOS version 6.4.1.

The following example illustrates how to enable and disable the telemetry debugging information.

Use this command to enable or disable gNMI server debugging logs with severity levels.

By default, this command is disabled, and the gNMI server debugging level in the disabled state is set to the Error level.

Configure Mode

Introduced in OcNOS version 6.4.1 and added the vrf (NAME|management) parameter in the OcNOS version 6.5.2.

The following example illustrates how to enable and disable the telemetry debug logs and their corresponding show output.

Use this command to enable the streaming telemetry and, upon configuration, to start the gNMI server. The gNMI server initiates listening for incoming gRPC connections on port 9339.

Use the no parameter of this command to disable the streaming telemetry, It will stop the gNMI server.

By default, the streaming-telemetry feature is disabled.

Configure mode

Introduced in OcNOS version 6.4.1 and added the vrf (NAME|management) parameter in the OcNOS version 6.5.2.

The following example illustrates how to enable the streaming telemetry on the default, management, and user-defined VRFs.

Use this command to display the streaming-telemetry details. of persistent (dial-out) and dynamic (dial-in) subscription connection details, including the POLL mode subscriptions.

The “show streaming-telemetry and all its sub-commands” also shows the max sensor-paths and minimum sample-interval for that platform.

None

Exec mode

This command was introduced in OcNOS version 6.5.2.

The following example displays the streaming telemetry details.

The below table explains the output fields.

Use this command to display streaming telemetry status in the running configuration.

None

Exec mode and Configuration Mode

This command was introduced in OcNOS version 6.4.1.

The following example shows the streaming telemetry status in the show running-config output.

Follow the below troubleshooting steps, to debug telemetry related issues:

Verify Collector (gnmic) Command Options: Verify the input parameters, such as the sensor path, prefix and origin “ipi:”.

Check the Encoding Method Compatibility: Check that the request conforms to the supported encoding methods.

Ensure Proper Connectivity: Validate the connectivity between the router and the remote management system. This involves verifying network settings, ports, firewalls, and any potential disruptions in communication.

Collector: If gnmic does not receive a response or not receiving expected response, restart the request using the “--log” option. If more verbose debug output is needed, consider adding the “--debug” option as well. The gnmic tool displays the possible cause for any error, which helps in debugging the issue.

gNMI Server: If the issue is on server side, follow the steps below to troubleshoot telemetry issues on the OcNOS target. Enable debug and verify the logs in /var/log/messages file.