beeline

package module
v0.2.2 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Sep 17, 2018 License: Apache-2.0 Imports: 6 Imported by: 0

README

Honeycomb Beeline for Go

Build Status GoDoc

This package and its subpackages contain bits of code to use to make your life easier when instrumenting a Go app to send events to Honeycomb. The wrappers here will collect a handful of useful fields about HTTP requests and SQL calls in addition to establishing easy patterns to augment this data as your application runs.

Documentation and examples are available via godoc.

Sign up for a Honeycomb trial to obtain a write key before starting.

Examples

For each of the wrappers, documentation is found in godoc rather than the github README files. Each package has a fully functional example in the godoc as well, to show all the pieces fit together.

Installation

Add the main package to your $GOPATH in the normal way: go get github.com/honeycombio/beeline-go. You'll add additional subpackages depending on the specifics of your application.

Setup

Regardless of which subpackages are used, there is a small amount of global configuration to add to your application's startup process. At the bare minimum, you must pass in your Team Write Key and identify a dataset name to authorize your code to send events to Honeycomb and tell it where to send events.

import "github.com/honeycombio/beeline-go"
...
func main() {
	beeline.Init(beeline.Config{
			WriteKey: "abcabc123123defdef456456",
			Dataset: "myapp",
		})
	...

Use

After initialization, the next step is to find the http.ListenAndServe call and add in the hnynethttp wrapper. This establishes the most basic instrumentation at the outermost layer of the call stack on a per-request basis.

	import "github.com/honeycombio/beeline-go/wrappers/hnynethttp"
	...
	http.ListenAndServe(":8080", hnynethttp.WrapHandler(muxer))

Now make a few requests in your web app and open up Honeycomb. You should see your new dataset! Open it to see the events you just sent.

Once this middleware wrapper is in place, there is a Honeycomb event in the request context available for use throughout the request's lifecycle. You could stop here and have very basic instrumentation, or continue to get additional context.

Example Questions

  • Which endpoints are the slowest?
BREAKDOWN: request.path
CALCULATE: P99(duration_ms)
FILTER: meta.type = http request
ORDER BY: P99(duration_ms) DESC
  • Which endpoints are the most frequently hit?
BREAKDOWN: request.path
CALCULATE: COUNT
FILTER: meta.type = http request
  • Which users are using the endpoint that I'd like to deprecate? (assuming we add a custom field with user.email)
BREAKDOWN: app.user.email
CALCULATE: COUNT
FILTER: request.url == <endpoint-url>

Example Event

Depending on which wrappers you use, events will have different fields. This example gives you a feel for what generated events will look like, though it may not exactly match what you'll see. All events have at least a timestamp, a duration, and the meta.type field.

{
    "Timestamp": "2018-03-20T00:47:25.339Z",
    "duration_ms": 0.809993,
    "meta.local_hostname": "cobbler.local",
    "meta.type": "http request",
    "mux.handler.name": "main.hello",
    "mux.handler.pattern": "/hello/",
    "mux.handler.type": "http.HandlerFunc",
    "request.content_length": 0,
    "request.header.user_agent": "curl/7.54.0",
    "request.host": "localhost:8080",
    "request.method": "GET",
    "request.path": "/hello/foo/bar",
    "request.proto": "HTTP/1.1",
    "request.remote_addr": "127.0.0.1",
    "response.status_code": 200,
    "trace.trace_id": "5279bdc7-fedc-483b-8e4f-a03b4dbb7f27"
}

Adding Additional Context

At any time (once the *http.Request is decorated with a Honeycomb event by the beeline) you can add additional custom fields to the event associated with this request.

beeline.AddField(req.Context(), "field_name", value)

By default additional fields you add are namespaced under app., so the field in the example above would appear in your event as app.field_name. This namespacing groups all your fields together to make them easy to find and examine.

These additional fields are your opportunity to add important and detailed context to your instrumentation. Put a timer around a section of code, add per- user information, include details about what it took to craft a response, and so on. It is expected that some fields will only be present on some requests. Error handlers are a great example of this; they will obviously only exist when an error has occurred.

It is common practice to add in these fields along the way as they are processed in different levels of middleware. For example, if you have an authentication middleware, it would add a field with the authenticated user's ID and name as soon as it resolves them. Later on in the call stack, you might add additional fields describing what the user is trying to achieve with this specific HTTP request.

Wrappers and Other Middleware

After the router has parsed the request, more fields specific to that router are available, such as the specific handler matched or any request parameters that might be attached to the URL. The wrappers for different HTTP routers handle this part differently. The next step is to add middleware or specific handler wrappers in additional places; instructions on how to do this are in each of the subpackages below.

Available HTTP wrappers:

Available DB wrappers:

  • hnysql (for database/sql)
  • hnysqlx (for github.com/jmoiron/sqlx)
  • pop (for github.com/gobuffalo/pop)

Other HTTP Frameworks

If your favorite framework isn't listed here, but supports middleware, look at the hnynethttp wrapper. Chances are, a phrase like "any middleware in the ecosystem that is also compatible with net/http can be used"(from go-chi) means that it expects a function that takes a http.Handler and returns a http.Handler, which is exactly what the WrapHandler function in hnynethttp does.

Try that out and see how far you can get with it and appropriate custom fields.

Optional Configuration

If you are using both an HTTP wrapper and a SQL package, make sure you pass the context from the *http.Request through to the SQL package using the various Context-enabled function calls. Doing so will tie the SQL calls back to specific HTTP requests and you'll get extra fields in your request event showing things like how much time was spent in the DB, as well as request IDs tying the separate events together so you can see exactly which DB calls were triggered by a given event.

For very high throughput services, you can send only a portion of the events flowing through your service by setting the SampleRate during initialization. This sample rate will send 1/n events, so a sample rate of 5 would send 20% of all events. For high throughput services, a sample rate of 100 is a good start.

Troubleshooting

There are two general approaches to finding out what's going wrong when the beeline isn't doing what you expect.

The events I'm generating don't contain the content I expect

Use the STDOUT flag in configuring the beeline. This will print events to the terminal instead of sending them to Honeycomb. This lets you quickly see what's getting sent and modify your code to get what you would like to see.

The events I'm sending aren't being accepted by Honeycomb

Use the Debug flag in configuring the beeline. This will print the responses that come back from Honeycomb to the terminal when sending events. These responses will have extra detail saying why events are being rejected (or that they are being accepted) by Honeycomb.

This beeline is also still young, so please reach out to support@honeycomb.io or ping us with the chat bubble on https://honeycomb.io for assistance.

Dependencies

The beeline is written with an eye towards a fairly recent set of dependent packages. Some of the dependencies have changed their API interface over the past few years (eg goji, uuid, net/http) so you may need to upgrade to get to a place that works.

  • go 1.9+ - the context package moved into the core library and is used extensively by the beeline to make events available to the call stack
  • github.com/google/uuid v0.2+ - the signature for NewRandom started returning UUID, error
  • github.com/goji/goji v2.0+ - they started using contexts in net/http instead of their own

Documentation

Overview

Package beeline aids adding instrumentation to go apps using Honeycomb.

Summary

This package and its subpackages contain bits of code to use to make your life easier when instrumenting a Go app to send events to Honeycomb. Most applications will use something out of the `wrappers` package and the `beeline` package.

The `beeline` package provides the entry point - initialization and the basic method to add fields to events.

The `trace` package offers more direct control over the generated events and how they connect together to form traces. It can be used if you need more functionality (eg asynchronous spans, other field naming standards, trace propagation).

The `propagation`, `sample`, and `timer` packages are used internally and not very interesting.

The `wrappers` package contains middleware to use with other existing packages such as HTTP routers (eg goji, gorilla, or just plain net/http) and SQL packages (including sqlx and pop).

Finally the `examples` package contains small example applications that use the various wrappers and the beeline.

Regardless of which subpackages are used, there is a small amount of global configuration to add to your application's startup process. At the bare minimum, you must pass in your team write key and identify a dataset name to authorize your code to send events to Honeycomb and tell it where to send events.

func main() {
  beeline.Init(&beeline.Config{
    WriteKey: "abcabc123123defdef456456",
    Dataset: "myapp",
  })
  ...

Once configured, use one of the subpackages to wrap HTTP handlers and SQL db objects.

Examples

There are runnable examples at https://github.com/honeycombio/beeline-go/tree/master/examples and examples of each wrapper in the godoc.

The most complete example is in `nethttp`; it covers - beeline initialization - using the net/http wrapper - creating additional spans for larger chunks of work - wrapping an outbound http call - modifying spans on the way out to scrub information - a custom sampling method

TODO create two comprehensive examples, one showing basic beeline use and the other the more exciting things you can do with direct access to the trace and span objects.

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func AddField

func AddField(ctx context.Context, key string, val interface{})

AddField allows you to add a single field to an event anywhere downstream of an instrumented request. After adding the appropriate middleware or wrapping a Handler, feel free to call AddField freely within your code. Pass it the context from the request (`r.Context()`) and the key and value you wish to add.This function is good for span-level data, eg timers or the arguments to a specific function call, etc. Fields added here are prefixed with `app.`

func AddFieldToTrace added in v0.2.0

func AddFieldToTrace(ctx context.Context, key string, val interface{})

AddFieldToTrace adds the field to both the currently active span and all other spans involved in this trace that occur within this process. Additionally, these fields are packaged up and passed along to downstream processes if they are also using a beeline. This function is good for adding context that is better scoped to the request than this specific unit of work, eg user IDs, globally relevant feature flags, errors, etc. Fields added here are prefixed with `app.`

func Close

func Close()

Close shuts down the beeline. Closing does not send any pending traces but does flush any pending libhoney events and blocks until they have been sent. It is optional to close the beeline, and prohibited to try and send an event after the beeline has been closed.

func Flush

func Flush(ctx context.Context)

Flush sends any pending events to Honeycomb. This is optional; events will be flushed on a timer otherwise. It is useful to flush before AWS Lambda functions finish to ensure events get sent before AWS freezes the function. Flush implicitly ends all currently active spans.

func Init

func Init(config Config)

Init intializes the honeycomb instrumentation library.

func StartSpan added in v0.2.0

func StartSpan(ctx context.Context, name string) (context.Context, *trace.Span)

StartSpan lets you start a new span as a child of an already instrumented handler. If there isn't an existing wrapped handler in the context when this is called, it will start a new trace. Spans automatically get a `duration_ms` field when they are ended; you should not explicitly set the duration. The name argument will be the primary way the span is identified in the trace view within Honeycomb. You get back a fresh context with the new span in it as well as the actual span that was just created. You should call `span.Send()` when the span should be sent (often in a defer immediately after creation). You should pass the returned context downstream.

Types

type Config

type Config struct {
	// Writekey is your Honeycomb authentication token, available from
	// https://ui.honeycomb.io/account. default: apikey-placeholder
	WriteKey string
	// Dataset is the name of the Honeycomb dataset to which events will be
	// sent. default: beeline-go
	Dataset string
	// Service Name identifies your application. While optional, setting this
	// field is extremely valuable when you instrument multiple services. If set
	// it will be added to all events as `service_name`
	ServiceName string
	// SamplRate is a positive integer indicating the rate at which to sample
	// events. Default sampling is at the trace level - entire traces will be
	// kept or dropped. default: 1 (meaning no sampling)
	SampleRate uint
	// SamplerHook is a function that will get run with the contents of each
	// event just before sending the event to Honeycomb. Register a function
	// with this config option to have manual control over sampling within the
	// beeline. The function should return true if the event should be kept and
	// false if it should be dropped.  If it should be kept, the returned
	// integer is the sample rate that has been applied. The SamplerHook
	// overrides the default sampler. Runs before the PresendHook.
	SamplerHook func(map[string]interface{}) (bool, int)
	// PresendHook is a function call that will get run with the contents of
	// each event just before sending them to Honeycomb. The function registered
	// here may mutate the map passed in to add, change, or drop fields from the
	// event before it gets sent to Honeycomb. Does not get invoked if the event
	// is going to be dropped because of sampling. Runs after the SamplerHook.
	PresendHook func(map[string]interface{})
	// APIHost is the hostname for the Honeycomb API server to which to send
	// this event. default: https://api.honeycomb.io/
	APIHost string
	// STDOUT when set to true will print events to STDOUT *instead* of sending
	// them to honeycomb; useful for development. default: false
	STDOUT bool
	// Mute when set to true will disable Honeycomb entirely; useful for tests
	// and CI. default: false
	Mute bool
	// Debug will emit verbose logging to STDOUT when true. If you're having
	// trouble getting the beeline to work, set this to true in a dev
	// environment.
	Debug bool
}

Config is the place where you configure your Honeycomb write key and dataset name. WriteKey is the only required field in order to actually send events to Honeycomb.

Directories

Path Synopsis
examples
pop
Package timer is a small convenience package for timing blocks of code.
Package timer is a small convenience package for timing blocks of code.
Package trace gives access to direct control over the span and trace objects used by the beeline.
Package trace gives access to direct control over the span and trace objects used by the beeline.
wrappers
hnygoji
Package hnygoji has Middleware to use with the Goji muxer.
Package hnygoji has Middleware to use with the Goji muxer.
hnygorilla
Package hnygorilla has Middleware to use with the Gorilla muxer.
Package hnygorilla has Middleware to use with the Gorilla muxer.
hnyhttprouter
Package hnyhttprouter has Middleware to use with the httprouter muxer.
Package hnyhttprouter has Middleware to use with the httprouter muxer.
hnynethttp
Package hnynethttp provides Honeycomb wrappers for net/http Handlers.
Package hnynethttp provides Honeycomb wrappers for net/http Handlers.
hnypop
Package hnypop wraps the gobuffalo/pop ORM.
Package hnypop wraps the gobuffalo/pop ORM.
hnysql
Package hnysql wraps `database.sql` to emit one Honeycomb event per DB call.
Package hnysql wraps `database.sql` to emit one Honeycomb event per DB call.
hnysqlx
Package hnysqlx wraps `jmoiron/sqlx` to emit one Honeycomb event per DB call.
Package hnysqlx wraps `jmoiron/sqlx` to emit one Honeycomb event per DB call.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL