Export Prometheus metrics when executing a job

[Baiquette]

Hello,

I'm running a PHP CronJob into a Kubernetes cluster and I would like to have the Prometheus Exporter during the execution.
The aim is to have an endpoint like http://localhost:2112/ read by Prometheus each X seconds.
I did some researchs by reading the documentation and I found 2 main solutions.
The best one seems to be the following:

In my .rr.yaml file:

# RR configuration version
version: "2.7"

rpc:
  listen: tcp://127.0.0.1:6001

service:
  some_service_1:
    command: "php my-script.php"
    process_num: 1
    exec_timeout: 0
    remain_after_exit: false
    restart_sec: 0

metrics:
  address: localhost:2112

And inside my-script.php:

<?php

require "vendor/autoload.php";

use Spiral\RoadRunner\Environment;
use Spiral\Goridge\RPC\RPC;
use Spiral\RoadRunner\Metrics\Metrics;

$rpc = RPC::create(Environment::fromGlobals()->getRPCAddress());

$metrics = new Metrics($rpc);
$metrics->add('app_metric_counter', 1);

This code works fine by starting the server with ./rr serve, and I can see the metrics on http://localhost:2112/

Now, I would like to stop Roadrunner to "delete" the container (because it is a Kubernetes CronJob).
I read that I can "Terminate" the service through RPC with:

// Kill the job
$manager = new Manager($rpc);
$manager->terminate('some_service_1');

But this code just terminate the service and not all the server.

Roadrunner has the ./rr reset command to reset the workers but I didn't find something like ./rr stop

Do you have an idea how to do it better?

[rustatian]

Hey @Baiquette 👋🏻. RR accepts SIGINT or SIGTERM syscalls to stop. On the stop, k8s will send these signals to the pod with RR, and RR will stop itself. So, you don't need any additional commands. After receiving SIGINT/SIGTERM RR will gracefully stop all active plugins. To manage the graceful timeout, you may set this configuration option: https://github.com/roadrunner-server/roadrunner/blob/master/.rr.yaml#L1596

EDIT: ./rr stop will land in the v2.10.3.

[Baiquette]

Okey thank you for your quick answer @rustatian
But in this case, I need to stop the pod.
The aim is to send a SIGINT/SIGTERM to RR from the PHP script once the execution done.
I would like to have something like:

// my script code...

// Stop the ./rr serve process
$rpc->stopRoadrunner();

Thank you :)

[rustatian]

Okey thank you for your quick answer @rustatian But in this case, I need to stop the pod. The aim is to send a SIGINT/SIGTERM to RR from the PHP script once the execution done. I would like to have something like:

// my script code...

// Stop the ./rr serve process
$rpc->stopRoadrunner();

Thank you :)

My pleasure 👍🏻

Yeah, this is possible if you know the RR's PID; however, I never tested such a scenario. And I think that you don't need an rpc to solve that; find the RR process via PHP and send the syscall directly. If you use a getmypid (or similar) function, you'll get a PHP process PID, not RRs.

It will be better to have a ./rr stop command.

[rustatian]

Could you please create a ticket for that with your case, and I'll include this feature in the upcoming v2.10.3 release?
EDIT: you may link this discussion w/o a profound explanation.

[Baiquette]

Ticket "Feature Request" created: #1162

Thank you for your reactivity and your feedback!

[rustatian]

Nice, thank you for your contribution 🤝