logo
search
Back to Pixie
EXPAND ALL
  • Home
WebsiteSlack Us

Using the API

This guide shows you how to get started with the Pixie API using one of our client libraries. For information on our gRPC API, see the reference docs.

Check out the following video for a live coding demo using the Go API:

Setup

  1. Install Pixie on your cluster using the install guides.
  2. Find your cluster's ID following the directions here.
  3. Create an API key following the directions here. An API Key is required when using the client API for running queries.

Install the client library

// We recommend using a Go modules project: https://blog.golang.org/using-go-modules
go get px.dev/pxapi

Import the library

// Import pixie go package and subpackages.
import (
    "context"
    "fmt"
    "io"
    "os"
    "px.dev/pxapi"
    "px.dev/pxapi/errdefs"
    "px.dev/pxapi/types"
)

Initialize a Pixie client and connect to a cluster

You'll need the API Key and Cluster ID created in the Setup section.

// Create a Pixie client.
ctx := context.Background()
client, err := pxapi.NewClient(ctx, pxapi.WithAPIKey(<YOUR_API_TOKEN_STRING>))
if err != nil {
    panic(err)
}
// Create a connection to the cluster.
vz, err := client.NewVizierClient(ctx, <YOUR_CLUSTER_ID_STRING>)
if err != nil {
    panic(err)
}

Run a PxL script and display the results

// Define a PxL script with out output table.
pxl := `
import px
df = px.DataFrame('http_events')
df = df[['upid', 'req_path', 'remote_addr', 'req_method']]
df = df.head(10)
px.display(df, 'http')
`
// Create TableMuxer to accept results table.
tm := &tableMux{}
// Execute the PxL script.
resultSet, err := vz.ExecuteScript(ctx, pxl, tm)
if err != nil && err != io.EOF {
    panic(err)
}
// Receive the PxL script results.
defer resultSet.Close()
if err := resultSet.Stream(); err != nil {
    fmt.Printf("Got error : %+v, while streaming\n", err)
}
// Satisfies the TableRecordHandler interface.
type tablePrinter struct{}
func (t *tablePrinter) HandleInit(ctx context.Context, metadata types.TableMetadata) error {
    return nil
}
func (t *tablePrinter) HandleRecord(ctx context.Context, r *types.Record) error {
    for _, d := range r.Data {
        fmt.Printf("%s ", d.String())
    }
    fmt.Printf("\n")
    return nil
}
func (t *tablePrinter) HandleDone(ctx context.Context) error {
    return nil
}
// Satisfies the TableMuxer interface.
type tableMux struct {
}
func (s *tableMux) AcceptTable(ctx context.Context, metadata types.TableMetadata) (pxapi.TableRecordHandler, error) {
    return &tablePrinter{}, nil
}

This PxL query returns two columns from the first 10 rows of the http_events table of data. For more information on how this PxL script was written, check out the PxL script tutorials.

Note that the API does not currently support running our open source px/ scripts by name. If you would like to run one of the px/ scripts that we include in the CLI or Live UI, you will need to copy the PxL script and pass it in as a string. Only PxL scripts with an empty Vis Spec will return results from the API.

If your PxL query has any compile errors, you will get a pixie.errors.PxLError. We recommend writing and debugging your PxL scripts using our Live UI, CLI tool.

Complete example source code

Below is the complete source code for the example above.

This basic example along with more advanced API examples can be found in the API folder in Pixie's GitHub repo.

package main
import (
    "context"
    "fmt"
    "io"
    "os"
    "px.dev/pxapi"
    "px.dev/pxapi/errdefs"
    "px.dev/pxapi/types"
)
// Define PxL script with one table output.
var (
    pxl = `
import px
df = px.DataFrame('http_events')
df = df[['upid', 'req_path', 'remote_addr', 'req_method']]
df = df.head(10)
px.display(df, 'http')
`
)
func main() {
    // Create a Pixie client.
    ctx := context.Background()
    client, err := pxapi.NewClient(ctx, pxapi.WithAPIKey(<YOUR_API_TOKEN_STRING>))
    if err != nil {
        panic(err)
    }
    // Create a connection to the cluster.
    vz, err := client.NewVizierClient(ctx, <YOUR_CLUSTER_ID_STRING>)
    if err != nil {
        panic(err)
    }
    // Create TableMuxer to accept results table.
    tm := &tableMux{}
    // Execute the PxL script.
    resultSet, err := vz.ExecuteScript(ctx, pxl, tm)
    if err != nil && err != io.EOF {
        panic(err)
    }
    // Receive the PxL script results.
    defer resultSet.Close()
    if err := resultSet.Stream(); err != nil {
        if errdefs.IsCompilationError(err) {
            fmt.Printf("Got compiler error: \n %s\n", err.Error())
        } else {
            fmt.Printf("Got error : %+v, while streaming\n", err)
        }
    }
    // Get the execution stats for the script execution.
    stats := resultSet.Stats()
    fmt.Printf("Execution Time: %v\n", stats.ExecutionTime)
    fmt.Printf("Bytes received: %v\n", stats.TotalBytes)
}
// Satisfies the TableRecordHandler interface.
type tablePrinter struct{}
func (t *tablePrinter) HandleInit(ctx context.Context, metadata types.TableMetadata) error {
    return nil
}
func (t *tablePrinter) HandleRecord(ctx context.Context, r *types.Record) error {
    for _, d := range r.Data {
        fmt.Printf("%s ", d.String())
    }
    fmt.Printf("\n")
    return nil
}
func (t *tablePrinter) HandleDone(ctx context.Context) error {
    return nil
}
// Satisfies the TableMuxer interface.
type tableMux struct {
}
func (s *tableMux) AcceptTable(ctx context.Context, metadata types.TableMetadata) (pxapi.TableRecordHandler, error) {
    return &tablePrinter{}, nil
}

End-to-End Encryption

Pixie offers end-to-end encryption for telemetry data requested by the API. For more information, see the FAQ.

Encryption is controlled by a flag during client creation. For additional info, see the Go or Python encryption examples.

This site uses cookies to provide you with a better user experience. By using Pixie, you consent to our use of cookies.