Oryx Helm chart impl nginx-hls-cdn

[suzp1984]

Background

ref https://github.com/ossrs/oryx/tree/main/scripts/nginx-hls-cdn#nginx-hls-cdn

The commit is the k8s helm chart implementation of nginx-hls-cdn. Inside a k8s cluster, deploy a oryx pod and a autoscale nginx pods that provide hls server to the frontend.

How to Install

1. download this repo;
2. `helm install oryx ./oryx-nginx-hls;
3. wait the containers download and pods deployed to k8s cluster;
4. check pod status by kubectl get pods, and k8s services by kubectl get services;
5. visit oryx front end by web browser: http://localhost:2080/mgmt/ (replace the localhost with the k8s service ip);
6. enable HLS high performance mode and HLS Low Latency if you want to, by this path in above frontend webui: System -> HLS.
7. push the video stream to the srs server;
8. play HLS stream by the k8s service ip of the nginx LoadBalancer. (check kubectl get services) e.g. ffplay -i http://127.0.0.1/live/livestream.m3u8

How to do benchmark in your local pc?

1. scripts to publish 50 streams to oryx or srs.

#!/bin/bash
for ((i=1;i<=50;i++)); 
do 
   # publish rtmp stream
   echo $i
   ffmpeg -nostdin -loglevel quiet -re -i trunk/doc/source.flv -c copy -f flv rtmp://localhost/live/livestream${i}?secret=fec73e1a68b84dc486240d9a9f952879 2>/dev/null &
done

the trunk/doc/source.flv is the test video source located at srs
the secret token is come from the oryx, check the oryx web UI.

2. script to pull 1000 hls streams from nginx cache.

#!/bin/bash
for ((i=1;i<=1000;i++)); 
do 
    # generate random number in range [1-50], represent which stream.
    r=`shuf -i 1-50 -n 1`
    echo $r
    ffmpeg -nostdin -loglevel quiet -i http://127.0.0.1/live/livestream${r}.m3u8 -f null - &
done

There is a k8s job to run this script in the cloud for the testing purpose.
https://gist.github.com/suzp1984/cc7d21e00b142c6f83a28d709b9954d3

apiVersion: batch/v1
kind: Job
metadata:
  name: ffmpeg-pull-test
spec:
  ttlSecondsAfterFinished: 120
  template:
    spec:
      containers:
      - name: ffmpeg
        image: "ossrs/oryx:v5.15.20"
        imagePullPolicy: IfNotPresent
        command:
        - bash
        - -c
        - |-
          echo "pull live streams by ffmpeg"
          echo "start 500 ffmpeg processes to pull streams"
          for ((i=1;i<=500;i++)); 
          do 
            # generate random number in range [1-50], represent which stream.
            r=`shuf -i 1-50 -n 1`
            echo $r
            ffmpeg -nostdin -loglevel quiet -i http://127.0.0.1/live/livestream${r}.m3u8 -f null - &
          done
          
          while true; do
            sleep 1
            if [[ $(ps aux |grep ffmpeg |grep -v grep |grep -v usr |grep -q ffmpeg || echo no) == no ]]; then
              echo "all ffmpeg process done"
              break
            fi
          done
          echo "complete ffmpeg job"
      restartPolicy: Never
  backoffLimit: 4

3. monitor pods metric
   kubectl top pod and kubectl top node