The signalfx adapter collects Istio metrics and trace spans and sends them to SignalFx.

This adapter supports the metric template and the tracespan template.

If sending trace spans, this adapter can make use of certain conventions in the tracespan format that is configured to send to this adapter. Here is an example tracespan spec that will work well:

kind: instance
  name: signalfx
  compiledTemplate: tracespan
    traceId: request.headers["x-b3-traceid"] | ""
    spanId: request.headers["x-b3-spanid"] | ""
    parentSpanId: request.headers["x-b3-parentspanid"] | ""
    # If the path contains query parameters, they will be split off and put into
    # tags such that the span name sent to SignalFx will consist only of the path
    # itself.
    spanName: request.path | "/"
    startTime: request.time
    endTime: response.time
    # If this is >=500, the span will get an 'error' tag
    httpStatusCode: response.code | 0
    clientSpan: context.reporter.kind == "outbound"
    # Span tags below that do not have comments are useful but optional and will
    # be passed to SignalFx unmodified. The tags that have comments are interpreted
    # in a special manner, but are still optional.
      # This is used to determine whether the span pertains to the client or
      # server side of the request.
      context.reporter.local: context.reporter.local
      # This gets put into the remoteEndpoint.ipv4 field
      destination.ip: destination.ip | ip("")
      # This gets flattened out to individual tags of the form
      # 'destination.labels.<key>: <value>'.
      destination.labels: destination.labels
      # This gets put into the field | "unknown"
      destination.namespace: destination.namespace | "unknown" | ""
      request.method: request.method | ""
      request.path: request.path | ""
      request.size: request.size | 0
      request.useragent: request.useragent | ""
      response.size: response.size | 0
      # This gets put into the field | "unknown"
      # This gets put into the localEndpoint.ipv4 field
      source.ip: source.ip | ip("")
      source.namespace: source.namespace | "unknown"
      # This gets flattened out to individual tags of the form
      # 'source.labels.<key>: <value>'.
      source.labels: source.labels
      source.version: source.labels["version"] | "unknown"


Configuration format for the signalfx adapter.


Required. The set of metrics to send to SignalFx. If an Istio metric is configured to be sent to this adapter, it must have a corresponding description here.


Optional. The URL of the SignalFx ingest server to use. Will default to the global ingest server if not specified.


Required. The access token for the SignalFx organization that should receive the metrics.


Optional. Specifies how frequently to send metrics to SignalFx. Metrics reported to this adapter are collected and reported as a timeseries. This will be rounded to the nearest second and rounded values less than one second are not valid. Defaults to 10 seconds if not specified.


Optional. If set to false, metrics won’t be sent (but trace spans will be sent, unless otherwise disabled).


Optional. If set to false, trace spans won’t be sent (but metrics will be sent, unless otherwise disabled).


Optional. The number of trace spans that the adapter will buffer before dropping them. This defaults to 1000 spans but can be configured higher if needed. An error message will be logged if spans are dropped.


Optional. The uniform probability ([0.0, 1.0]) that a given span gets sampled if its parent was not already sampled. Child spans will always be sampled if their parent is. If not provided, defaults to sending all spans.


Describes what metrics should be sent to SignalFx and in what form.


Required. The name of the metric as it is sent to the adapter. In Kubernetes this is of the form “<name>.metric.<namespace>” where “<name>” is the name field of the metric resource, and “<namespace>” is the namespace of the metric resource.


The metric type of the metric


Describes what kind of metric this is.


None is the default and is invalid


Values with the same set of dimensions will be added together as a continuously incrementing value.


A histogram distribution. This will result in several metrics emitted for each unique set of dimensions.