Apstra Telemetry Streaming Plugin for Telegraf

Juniper Apstra 5.1 introduces an Apstra telemetry streaming input plugin for Telegraf. Supported and maintained as an external input plugin for Telegraf, the Apstra streaming plugin runs through Telegraf’s execd plugin. Previously, the Apstra plugin’s source code was compiled within Telegraf, making it dependent on custom Telegraf forks for each Telegraf release.

In order to address version compatibility issues and improve the user experience, Juniper Apstra 5.1 has transitioned to an external plugin model for Apstra telemetry streaming. The plugin is a statically linked executable binary file. Building the plugin as an external plugin model means that it is deployed and updated independently of Telegraf’s updates, enabling users to integrate the Apstra streaming plugin into most current monitoring stacks without compatibility issues. Additionally, Telegraf provides detailed documentation for external plugins, with configuration and deployment information.

As stated in Telegraf’s official execd documentation, the input plugin is part of Telegraf’s extensible architecture, which lets Telegraf execute external scripts for data collection as long-running processes. In the case of Asptra’s streaming input plugin, execd lets Telegraf interface with the binary to run Apstra’s telemetry streaming scripts.

The Apstra plugin streams the following data types:

• perfmon (performance data)

• alerts

• events

The available execd configuration options and their descriptions are available in the official documentation.

• Telegraf Metrics in the Apstra Plugin
• Perfmon Metrics
• Alerts
• Events

Telegraf Metrics in the Apstra Plugin

The following sections list the various Telegraf metrics that the Apstra plugin streams. Telegraf uses these metrics to model data during processin. These metrics contain four main components:

• Measurement name: Description and namespace for the metric.

• Tags: Key/Value string pairs, used to identify the metric.

• Fields: Key/Value pairs that are typed and usually contain the metric data.

• Timestamp: Date and time associated with the fields.

For output to InfluxDB, Telegraf converts metrics to InfluxDB line protocol. For more information, see Telegraf's official documentation at https://docs.influxdata.com/telegraf/v1/metrics/.

Consider the following example:

When applied to an Apstra Telegraf plugin metric like alert_streaming for example, the line protocol output might look something like the following:

Measurement: alert_bgp_neighbor_mismatch

Tags: addr_family, blueprint, device...etc

Fields: expected_state, actual_state, status, sequence_number

Timestamp: 1738596731833786640

Note: Some metrics, such as certain Events metrics, don't stream any field metric data.

The Apstra Telegraf plugin for Juniper Apstra 5.1 is not backwards compatible with the previous (deprecated) version. The following data measurements are not supported in this (5.1) version.

Removed:

• event_traffic

• alert_counters

• alert_headroom

• alert_config_mismatch_alert

• generic

The following new fields and tags are supported:

• For all streamed messages, except alert_streaming and event_streaming, the blueprint_id tag was added.

• For alert_probe measurement, the predefined_probe_name tag was added.

• For interface_counters measurement, the delta_microseconds field was added.

The following changes were made to existing fields and tags:

• Previously, for all event and alert measurements, all data fields streamed as fields. Starting with the Apstra 5.1.0 plugin, the subset of fields is moved to tags. For detailed information, see the Apstra Plugin Metrics section.

• Previously, the Apstra Telegraf plugin divided interface_counters measurement values into delta_seconds field that caused incorrect values. The following list of fields returns the correct per second values.

  • tx_bps

  • tx_unicast_pps

  • tx_broadcast_pps

  • tx_multicast_pps

  • tx_error_pps

  • tx_discard_pps

  • rx_bps

  • rx_unicast_pps

  • rx_broadscase_pps

  • rx_multicast_pps

  • rx_error_pps

  • rx_discard_pps

• The following streamed sets of tags and fields was changed for the probe_message messages:

  • The probe_id, probe_label, and stage_name fields are streamed as tags.

  • The format of the property field was changed. This field is divided into multiple fields in the property.<key> format, where the <key> is property key name and the value of the field is the property value.

  • The value field returns the correct data type depending on the processor output. The value field is also divided into multiple fields for the following processor outputs, according to the mapping.

    For the Active Floodlist processor, the following fields are streamed:

    • value.state

    • value.system_id

    • value.vni

    • value.vtep

  For EVPN Type 3 processor, the following fields are streamed:

  • value.state

  • value.vni

  • value.nexthop

  • value.rd

  • value.rt

  For EVPN Type 5 processor, the following fields are streamed:

  • value.state

  • value.system_id

  • value.af

  • value.subnet

  • value.nexthop

  • value.rd

  • value.rt

  For Interface Utilization processor the following fields are streamed:

  • value.tx_utilization

  • value.rx_utilization

  • value.tx_unicast_pps

  • value.tx_broadcast_pps

  • value.tx_multicast_pps

  • value.tx_bps

  • value.tx_error_pps

  • value.tx_discard_pps

  • value.rx_unicast_pps

  • value.rx_broadcast_pps

  • value.rx_multicast_pps

  • value.rx_bps

  • value.rx_error_pps

  • value.rx_discard_pps

  • value.alignment_errors_per_second

  • value.fcs_errors_per_second

  • value.symbol_errors_per_second

  • value.runts_per_second

  • value.giants_per_second

  For System Interface Utilization processor, the following fields are streamed:

  • value.aggregate_tx_bps

  • value.aggregate_rx_bps

  • value.max_ifc_tx_utilization

  • value.max_ifc_rx_utilization

  • value.aggregate_tx_utilization

  • value.aggregate_rx_utilization