[improve][stats] Enhance bookie metrics to support URL parameter filtering

[zwOvO]

Motivation

Allow filtering of metrics using URL parameters, for example:

/metrics?m_regex=bookkeeper_server_thread_executor_queue
/metrics?m_regex=.*_thread_executor_queue

Master Issue: #4143

[hangc0276]

@asafm Would you please help take a look at this improvement? Thanks.

[asafm]

@hangc0276 /metrics should be Prometheus compatible.
If you look at the Prometheus Java client, setting up a server, exposing /metrics you'll see it only supports specifying the metric names, using the query param name[].

    protected static Set<String> parseQuery(String query) throws IOException {
        Set<String> names = new HashSet<String>();
        if (query != null) {
            String[] pairs = query.split("&");
            for (String pair : pairs) {
                int idx = pair.indexOf("=");
                if (idx != -1 && URLDecoder.decode(pair.substring(0, idx), "UTF-8").equals("name[]")) {
                    names.add(URLDecoder.decode(pair.substring(idx + 1), "UTF-8"));
                }
            }
        }
        return names;
    }

Besides that, I presume the idea is that BookKeeper users will transition into OpenTelemetry which is already available. For now, even OpenTelemetry Java SDK doesn't support filtering YET.
I added a proposal to OTel specifications and it was approved, and hopefully in the following months it will be implemented.

For now I recommend not adding it, and using OTel collector to scrape /metrics and filter out any unwanted metrics.

[zwOvO]

OK, Now I am filtering JVM metrics by writing a Python server proxy
——————

#!/bin/python

import SimpleHTTPServer
import SocketServer
import subprocess
import requests

class CustomHandler(SimpleHTTPServer.SimpleHTTPRequestHandler):
    def do_GET(self):
        if self.path == '/metrics':
            
            response = requests.get('http://localhost:8080/metrics/')
            output = '\n'.join(line for line in response.text.split('\n') if 'jvm' in line)
            self.send_response(200)
            self.send_header('Content-type', 'text/plain')
            self.end_headers()
            self.wfile.write(output)

PORT = 9001
httpd = SocketServer.TCPServer(("", PORT), CustomHandler)
print "Server running on port", PORT
httpd.serve_forever()

[asafm]

OK, Now I am filtering JVM metrics by writing a Python server proxy ——————

#!/bin/python

import SimpleHTTPServer
import SocketServer
import subprocess
import requests

class CustomHandler(SimpleHTTPServer.SimpleHTTPRequestHandler):
    def do_GET(self):
        if self.path == '/metrics':
            
            response = requests.get('http://localhost:8080/metrics/')
            output = '\n'.join(line for line in response.text.split('\n') if 'jvm' in line)
            self.send_response(200)
            self.send_header('Content-type', 'text/plain')
            self.end_headers()
            self.wfile.write(output)

PORT = 9001
httpd = SocketServer.TCPServer(("", PORT), CustomHandler)
print "Server running on port", PORT
httpd.serve_forever()

You can just use OTel Collector. It has built-in functionality to do just that.