HLS: How to use NGINX to setup a HLS delivery cluster.

[winlinvip]

It works like this:

FFmpeg/OBS ---RTMP--> SRS Origin Server ---HLS--> NGINX Servers --> Players of Visitors

SRS is serve as an Origin server, to covert RTMP to HLS, and feed all NGINX Servers.

There are many NGINX Servers, to serve a set of Players. For example, if we got 10 NGINX Servers, each bandwidth is 5Gbps, then we can serve 50Gbps streams. That is, if each stream is 1Mbps, we could serve about 50k visitors.

Please read wiki:

• CN: https://github.com/ossrs/srs/wiki/v4_CN_NginxForHLS
• EN: https://github.com/ossrs/srs/wiki/v4_EN_NginxForHLS

[winlinvip]

Nginx Cache Control

If ignore the cache control header, NGINX won't follow the upstream, literally should NOT set it:

proxy_ignore_headers Cache-Control Expires;

By default, Nginx uses cache-control to overwrite the proxy_cache_valid, please read this post.

So, we should set the cache-control by upstream server, SRS or NGINX:

# For HLS delivery
location ~ /.+/.*\.(m3u8)$ {
  root /usr/local/lighthouse/softwares/srs-terraform/mgmt/containers/objs/nginx/html;
  add_header Cache-Control "public, max-age=10";
}
location ~ /.+/.*\.(ts)$ {
  root /usr/local/lighthouse/softwares/srs-terraform/mgmt/containers/objs/nginx/html;
  add_header Cache-Control "public, max-age=86400";
}

Note that SRS does not set the Cache-Control, see #2991

[winlinvip]

Fixed.