rtsp-stream

command module
v1.4.1 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: May 5, 2019 License: MIT Imports: 8 Imported by: 0

README

Go Report Card Maintainability GitHub last commit GitHub release

rtsp-stream is an easy to use out of box solution that can be integrated into existing systems resolving the problem of not being able to play rtsp stream natively in browsers.

Table of contents

How does it work

It converts RTSP streams into HLS based on traffic. The idea behind this is that the application should not transcode anything until someone is actually watching the stream. This can help with network bottlenecks in systems where there are a lot of cameras installed.

There's a running go routine in the background that checks if a stream is being active or not. If it's not the transcoding stops until the next request for that stream.

Authentication

The application offers different ways for authentication. There are situations when you can get away with no authentication, just trusting requests because they are from reliable sources or just because they know how to use the API. In other cases, production cases, you definitely want to protect the service. This application was not written to handle users and logins, so authentication is as lightweight as possible.

No Authentication

By default there is no authentication what so ever. This can be useful if you have private subnets where there is no real way to reach the service from the internet. (So every request is kind of trusted.) Also works great if you just wanna try it out, maybe for home use.

JWT Authentication

You can use shared key JWT authentication for the service.

The service itself does not create any tokens, but your authentication service can create. After it's created it can be validated in the transcoder using the same secret / keys. It is the easiest way to integrate into existing systems.

The following environment variables are available for this setup:

Env variable Description Default Type
RTSP_STREAM_AUTH_JWT_ENABLED Indicates if the service should use the JWT authentication for the requests false bool
RTPS_STREAM_AUTH_JWT_SECRET The secret used for creating the JWT tokens macilaci string
RTSP_STREAM_AUTH_JWT_PUB_PATH Path to the public shared RSA key. /key.pub string
RTSP_STREAM_AUTH_JWT_METHOD Can be secret or rsa. Changes how the application does the JWT verification. secret string

You won't need the private key for it because no signing happens in this application.

Easy API

There are 2 main endpoints to call:

POST /start

Starts the transcoding of the given stream. You have to pass URI format with rtsp procotol. The respond should be considered the subpath for the video player to call. So if your applicaiton is myapp.com then you should call myapp.com/stream/host/index.m3u8 in your video player. The reason for this is to remain flexible regarding useability.

Requires payload:

{ "uri": "rtsp://username:password@host" }

Response:

{ "uri": "/stream/host/index.m3u8" }

GET /stream/host/*file

Simple static file serving which is used when fetching chunks of HLS. This will be called by the client (browser) to fetch the chunks of the stream based on the given index.m3u8


GET /list

This (kind of a debug) endpoint is used to list the streams in the system. Since the application does not handle users, it does not handle permissions obviously. You might not want everyone to be able to list the streams available in the system. But if you do, you can use this. You just have to enable it via env variable.

Response:

[
    {
        "running": true,
        "uri": "/stream/185.180.88.98-streaming-channels-101/index.m3u8"
    }
]

Configuration

You can configure the following settings in the application with environment variables:

Env variable Description Default Type
RTSP_STREAM_CLEANUP_TIME Time period for the cleanup process info on format here 2m0s string
RTSP_STREAM_STORE_DIR Sub directory to store the video chunks ./videos string
RTSP_STREAM_KEEP_FILES Option to keep the chunks for the stream being transcoded false bool

The project uses Lumberjack for the log rotation of the ffmpeg transcoding processes.

Env variable Description Default Type
RTSP_STREAM_PROCESS_LOGGING_ENABLED Indicates if logging of transcoding ffmpeg processes is enabled or not false bool
RTSP_STREAM_PROCESS_LOGGING_DIR Describes the directory where the transcoding logs are stored /var/log/rtsp-stream string
RTSP_STREAM_PROCESS_LOGGING_MAX_SIZE Maximum size of each log file in megabytes 500 integer
RTSP_STREAM_PROCESS_LOGGING_MAX_AGE Maximum number of days that we store a given log file. 7 integer
RTSP_STREAM_PROCESS_LOGGING_MAX_BACKUPS Maximum number of old log files to retain 3 integer
RTSP_STREAM_PROCESS_LOGGING_COMPRESS Option to compress the rotated log file or not true bool

Env variable Description Default Type
RTSP_STREAM_PORT Port where the application listens 8080 integer
RTSP_STREAM_DEBUG Turns on / off debug logging false bool
RTSP_STREAM_LIST_ENDPOINT Turns on / off the /list endpoint false bool

By default all origin is allowed to make requests to the server, but you might want to configure it for security reasons.

Env variable Description Default Type
RTSP_STREAM_CORS_ENABLED Indicates if cors should be handled as configured or as default (everything allowed) false bool
RTSP_STREAM_CORS_ALLOWED_ORIGIN A list of origins a cross-domain request can be executed from []string
RTSP_STREAM_CORS_ALLOW_CREDENTIALS Indicates whether the request can include user credentials like cookies, HTTP authentication or client side SSL certificates false bool
RTSP_STREAM_CORS_MAX_AGE Indicates how long (in seconds) the results of a preflight request can be cached. 0 integer

Run with Docker

The application has an offical docker repository at dockerhub, therefore you can easily run it with simple commands:

docker run -p 80:8080 roverr/rtsp-stream:1

or you can build it yourself using the source code.

UI

You can use the included UI for handling the streams. The UI is not a compact solution right now, but it gets the job done.

Running it with docker:

docker run -p 80:80 -p 8080:8080 roverr/rtsp-stream:1-management

If you decide to use the management image, you should know that port 80 is flexible, you can set it to whatever you prefer, but 8080 is currently burnt into the UI as the ultimate port of the backend.

You should expect something like this:

Logs

With the UI solution the following files are created in /var/log:

  • rtsp-stream-ui.err.log
  • rtsp-stream-ui.out.log
  • rtsp-stream.err.log
  • rtsp-stream.out.log

Proven players

The following list of players has been already tried out in production environment using this backend:

Coming soon features

✅ - Done

🤷‍♂️ - Needs more labour

  • ✅ Proper logging - File logging for the output of ffmpeg with the option of rotating file log
  • ✅ Improved cleanup - Unused streams should be removed from the system after a while
  • 🤷‍♂️ API improvements - Delete endpoint for streams so clients can remove streams whenever they would like to
  • ✅ Authentication layer - More options for creating authentication within the service

Documentation

The Go Gopher

There is no documentation for this package.

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL