Skip to content

Latest commit

 

History

History
192 lines (136 loc) · 7.99 KB

integration.md

File metadata and controls

192 lines (136 loc) · 7.99 KB
Raw

Integration data model

The integration JSON payload contains the complete information sample that an integration sends through the New Relic Infrastructure Agent, at a given moment.

This document describes the basic structure and elements of the integration payload, as well as the basic API by means of simple examples.

For detailed information about the integration package, please refer to the Integration package GoDoc page

Building an integration

To build an integration with the default configuration, you have to invoke the integration.New function with the integration name and version as argument:

payload, err := integration.New("my-integration", "1.0")

The above invocation will return an integration with the default configuration:

  • The final integration JSON payload is sent to the standard output.
  • The logging messages are submitted to the standard error (with INFO level).
  • A persistent Storer is set, whose contents will be stored in a file whose path can be constructed as <OS temp dir>/nr-integrations/<integration name>.json, whith a default 1-minute Time To Live.
  • Configuration specified in the default arguments.

The integration.New function accepts, as a variable number of arguments, diverse configuration options. For example, the following code would create an integration which logs data to a Logrus logger implementation. In addition, the output JSON payload is written to a file called output.json instead of the default standard output.

payloadFile, _ := os.Create("output.json")

payload, err := integration.New("my-integration-data", "1.0",
        integration.Logger(logrus.New()),
        integration.Writer(payloadFile),
    )

For more details, check the documentation of the Option interface implementations.

You can safely add data from different concurrent threads since the sdk is thread safe.

Integration structure elements

An integration JSON payload contains data from multiple entities. Each entity stores information about metrics, inventory and events.

The rest of this section describes all the concepts that are part of the integration JSON payload, as well as their basic composition through the GoSDK v3 API.

Entity

An entity represents a monitoring target (e.g. a host). Since GoSDK v3, a single JSON payload can handle data from multiple entities (for example, the local host that runs the Infrastructure Agent + some remote or virtual hosts that are being monitored by the integration).

To instantiate a new entity, you can use the LocalEntity or Entity functions from the Integration type:

payload, err := integration.New("my-integration-data", "1.0")

// Creates a local entity
localhost := payload.LocalEntity()

// Creates a "remote" entity, given an entity name and a namespace
container, err := payload.Entity("my-cloud-resource", "my-namespace")

Each entity has three sections: metrics, inventory, and events, which will be explained in the following subsections.

Metrics

Metrics are quantifiable measurements associated to a given entity (e.g. percentage of CPU, or requests/second). There are four type of metrics:

  • GAUGE: an absolute, spot value, such as percentage of used CPU or free system memory.
  • RATE: a value that represents a measured quantity on a time period. For example, transferred bytes per second.
  • PRATE: version of RATE that only allows positive values.
  • DELTA: a value that represents the variation of a quantity since the last sample. For example, number of new connections since the last metric sample was set.
  • PDELTA: version of DELTA that only allows positive values.
  • ATTRIBUTE: a string value (e.g. "high" or "stopped")

Metrics are grouped in a key-value map (called metrics set). Every entity can control several, different metric sets. The Entity type provides functions to create metric sets (as well as for inventory and events):

myHost := payload.LocalEntity()

cpu := myHost.NewMetricSet("CpuSample")
err = cpu.SetMetric("cpuPercent", 75.0, metric.GAUGE)

RATE, PRATE, DELTA and PDELTA require to belong to at least 1 attribute.

As they are flushed to disk this attribute is used to "namespace" the metrics on the set so they don't collide with others with the same name.

So NewMetricSet provides an optional list of metric.Attribute arguments. There's a constructor function that comes handy to create one: metric.Attr.

The attributes provided on the NewMetricSet constructor are also added as usual attribute metrics.

If no Attribute is provided to NewMetricSet, an error value will be returned when calling SetMetric for a RATE, PRATE, DELTA or PDELTA.

disk := myHost.NewMetricSet("DiskSample", metric.Attr("diskStatus", "OK"))
err1 = disk.SetMetric("readsPerSecond", 12, metric.RATE)
err2 = disk.SetMetric("readBytes", 134, metric.DELTA)

The above example creates two metric sets for a same entity. The NewMetricSet function accepts the name of the metric set as a parameter, then the SetMetric function requires the name, value as well as the type of the metric.

Please refer to the Metrics GoDoc for a detailed description of the metrics API.

Inventory

Inventory provides track of a set of available items, as well as some associated data to them. For example, the installed software components, with their version number.

Inventory items are stored as a 2-level map, where the first-level key represents the inventory item and the second-level key plus its associated value represents a given aspect of such inventory item.

To set the inventory items and associated values, the Entity type provides the SetInventoryItem(<item name>, <key name>, <value>). For example, the following code keeps track of the items inside a remote entity named "fridge" in the "beach-home" namespace, and describes different characteristics for some of the items inside the fridge:

payload, _ := integration.New("my-integration-data", "1.0")

food, _ := payload.Entity("fridge", "beach-home")

food.SetInventoryItem("beans", "weight", 125.0)
food.SetInventoryItem("beans", "brand", "Heinz")
food.SetInventoryItem("beer", "size", 1000)
food.SetInventoryItem("beer", "type", "dark")
food.SetInventoryItem("coffee", "decaf", "yes")

payload.Publish()

The above code would insert the next inventory data into the integration JSON payload:

"inventory": {
    "beans": {
      "brand": "Heinz",
      "weight": 125
    },
    "beer": {
      "size": 1000,
      "type": "dark"
    },
    "coffee": {
      "decaf": "yes"
    }
  },

Please refer to the Inventory GoDoc for a detailed description of the inventory API.

Events

Events describe meaningful things that happen at a given moment (e.g. a host has been started, a package has been removed, a configuration property has changed...).

An event has two fields: summary and category. Category is a simple keyword to group events. Summary is a human-readable descriptive message. An event is created by the event.New function, and added by means of the AddEvent function of the Entity type.

For example, the following code would add two different events from different categories:

myHost := payload.LocalEntity()

myHost.AddEvent(event.New("/etc/httpd.conf configuration file has changed", "config"))
myHost.AddEvent(event.New("Service httpd has been restarted", "services"))

Please refer to the Events GoDoc for a detailed description of the events API.