OSSRS Case HTTP-FLV Cluster

OSSRS is a feature rich, high performance, live streaming solution. For more information, visit:

• ZH Wiki
• EN Wiki

This post setup a working Clustered HTTP-FLV.

Server arrangement

We will use pure Docker to deploy the server instances, so the configuration can be shared by them. As cross-server container communication is impossible, so the coworker address must be specified explicitly.

Configuration

Configuration files will be put at e.g., ${PWD}/conf which means you need to provide a conf directory under wherever you want.

origin server

Place the config at ${PWD}/conf/http.flv.live.conf :

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

listen              1935;
max_connections     1000;
daemon              off;
srs_log_tank        console;

http_server {
    enabled         on;
    listen          8080;
    dir             ./objs/nginx/html;
}

http_api {
    enabled on;
    listen  9090;
}

vhost __defaultVhost__ {
    http_remux {
        enabled     on;
        mount       [vhost]/[app]/[stream].flv;
        hstrs       on;
    }

    cluster {
        mode           local;
        origin_cluster on;
        coworkers      origin-coworker-1:9090;
    }
}

Ports are used for:

• 1935 RTMP
• 9090 cluster communication
• 8080 HTTP-FLV

edge server

Place the config at ${PWD}/conf/http.flv.live.edge1.conf :

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

listen              1935;
max_connections     1000;
daemon              off;
srs_log_tank        console;

http_server {
    enabled         on;
    listen          8080;
    dir             ./objs/nginx/html;
}
vhost __defaultVhost__ {
    mode remote;
    origin origin-server-1 origin-server-2;
    http_remux {
        enabled     on;
        mount       [vhost]/[app]/[stream].flv;
        hstrs       on;
    }
}

Ports are used for:

• 1935 RTMP
• 8080 HTTP-FLV

Run

origin server

On serverA :

replace origin-coworker-1 with 192.168.0.3, as serverB deploys a origin server

On serverB :

replace origin-coworker-1 with 192.168.0.2, as serverA deploys a origin server

注意:

• 在没有容器集群的情况下, Cluster 的 coworker 地址必须明确指定. 如果使用hostname, 则 edge server 会通过 hostname 对流进行重定向, 当没有集群来支持容器跨节点访问时, 就会无法访问.
• 在有容器集群的情况下, 同样可以使用 hostname 或者 service name

Then start the origin server on serverA and serverB with:

1
2
3
4

docker run -d --name srs-origin-1 \
    -p 1935:1935 -p 9090:9090 -p 8081:8080 \
    -v `pwd`/conf/http.flv.live.conf:/usr/local/srs/conf/srs.conf \
    ossrs/srs:3

edge server

Start on serverA and serverB with:

1
2
3
4
5
6

docker run -d --name srs-edge-1 \
    -p 19351:1935 -p 8082:8080 \
    --add-host=origin-server-1:192.168.0.2 \
    --add-host=origin-server-2:192.168.0.3 \
    -v `pwd`/conf/http.flv.live.edge1.conf:/usr/local/srs/conf/srs.conf \
    ossrs/srs:3

Result

In case we are

• publishing a RTMP stream to srs-edge-1 on serverB (ie., 192.168.0.3:19351),
• and that edge server forwarded to srs-origin-1 on serverA

Then the stream services’ availability would be:

Conclusion

For more robust performance, both publish and play stream via edge server is always recommended.

Reason: edge server is designed to auto fail-over and load balance the upstream Clustered Origin Server.

References

• Origin Cluster
• HTTP-FLV
• Docs