splunk-persistentconn implements the protocol that splunk core uses to communicate with persistent script that's used to power REST endpoints of an app
Usage example:
package main
import (
"fmt"
"net/http"
"time"
conn "github.com/DrakeW/splunk-persistentconn"
)
func main() {
// server initialization
server := conn.NewServer()
// register endpoint using either static path or path pattern
server.Handle("entity/:id/data", func(req conn.Request) (conn.Response, error) {
return conn.Response{
StatusCode: http.StatusOK,
Body: fmt.Sprintf("hello world %s", req.Params["id"]),
}, nil
}, "GET", "POST")
// starts server
server.Run()
}
An example app is located in the examples
directory.
This package is built to take advantage of the concurrent processing capability provided by goroutines for each handler registered. However, since Splunk core sends requests to the server process in a synchronous manner, only one goroutine is spawned at a time therefore currently there's no performance boost.
Based on rough benchmark with the same 5000 non-blocking requests (where each request takes ~0.5s to finish) hitting both the python REST endpoint based on PersistentServerConnectionApplication
and the Go REST endpoint using the splunk-persistentconn
package, performance is roughly the same.
This package is not officially supported by Splunk. Please use with caution and feel free to open issue or PR for any kind of bug report or feature enhancement.