fake-firehose/readme.md
2022-12-28 01:27:21 +00:00

10 KiB

Fake Firehose

This project is basically a shell/bash/text frontend for fakerelay

It allows instances to fill their federated timelines from other instances that have public timelines.

You can find the fakefirehose author at @raynor@raynor.haus

How to run it

In the config folder there are three files

  • domains-federated
  • domains-local
  • hashtags

If you want the full on public feed from an instance, put it in the domains-federated file, one domain per line.

If you only want the local feed from an instance, put it on the domains-local file, one domain per line.

If you want to follow a hash tag you either either add a hashtag after an instance in domains-federated or domains-local

For example: if in domains-fedarated you put mastodon.social #JohnMastodon a stream will open to watch for the hashtag #JohnMastodon on the public stream from mastodon.social

Another example: if in domains-local you put infosec.exchange #hacker a stream will open to watch for the hashtag #hacker on the local stream from infosec.exchange

Docker

To run it in docker -- recommended

  1. Make sure you have docker installed.
  2. From your shell: create a directory, it is recommended that you give it a relevant name
  3. Go into that directory and use git clone https://github.com/raynormast/fake-firehose.git
  4. Go into the created directory: cd fake-firehose
  5. docker build -t fakefirehose .
  6. Edit your docker-compose.yml file as needed. The biggest thing is to watch the volumes. It is highly recommended that you keep your data directory in the parent directory, and NOT the directory the git repo is in.
  7. Edit your .env.production file. The file is fairly well commented.
  8. Run docker compose -f docker-compose.yml

The entire thing should look something like:

cd ~
mkdir MastodonFireHose
cd MastodonFirehose
git pull https://github.com/raynormast/fake-firehose.git
cd fake-firehose
docker build -t fakefirehose .
# Edit your docker-compose and .env.production here
sudo docker compose -f docker-compose.yml up -d

Configuration

tl;dr

Your ./config folder has three sample files, after editing you should have the following three files:

domains-federated
domains-local
hashtags

In each file, comments begin with ## not the tradional single #.

The syntax is the same for the domains files:

## Follow full timeline
mastodon.instance

## Follow these hashtags from the timeline
mastodon.instance #mastodon #relay

The files are well commented.

domains-federated file

This file has the full federated feeds of any instances you want fed to fakerelay.

Each line of the file should have the domain name of an instance whose federated timeline you want to follow. I.e.,

raynor.haus
infosec.exchange

This can generate a LOT of posts if you choose a large instance.

For example, if you use mastodon.social or mas.to you can expect your server to fall behind. mastodon.social generates 50,000 - 200,000 posts on the federated timeline per day.

It is recommended that you only use this file to:

  • follow hashtags
  • follow instances with small federated timelines, with content you want in yours

domains-federated hashtags

The one time to use the federated timeline is to catch most posts with a specific hashtag.

Every word after after an instance domain is a hashtag to relay.

Example:

mastodon.social fediblock fediverse mastodev mastoadmin

Will only return posts from the mastodon.social federated feed with hashtags of #fediblock, #fediverse, #mastodev, and #mastoadmin.

The # is optional -- it is accepted simply to make the file more intuitive.

domains-local file

This file is identical to the domains-federated file except that it only recieves posts created on that instance (the local timeline).

It is possible to keep up with the larger instances, such as mastodon.social if you only look at the local timeline.

hashtags file

If you put ANY hashtags in here a stream will be opened for every host in the domains-federated and domains-local file.

It's purpose is for people or instances that want to find nearly every post with a particular hashtag

It can very quickly open up a lot of curl streams

Example

domains-federated content:

mastodon.social
mas.to

domains-local content:

aus.social
mastodon.nz

hashtags content:

JohnMastodon
Mastodon

will result in the following streams all opening:

https://mastodon.social/api/v1/streaming/public
https://mas.to/api/v1/streaming/public
https://aus.social/api/v1/streaming/public/local
https://mastodon.nz/api/v1/streaming/public/local
https://mastodon.social/api/v1/streaming/hashtag?tag=JohnMastodon
https://mas.to/api/v1/streaming/hashtag?tag=JohnMastodon
https://aus.social/api/v1/streaming/hashtag?tag=JohnMastodon
https://mastodon.nz/api/v1/streaming/hashtag?tag=JohnMastodon
https://mastodon.social/api/v1/streaming/hashtag?tag=Mastodon
https://mas.to/api/v1/streaming/hashtag?tag=Mastodon
https://aus.social/api/v1/streaming/hashtag?tag=Mastodon
https://mastodon.nz/api/v1/streaming/hashtag?tag=Mastodon

If you had a total of 5 lines in domains-federated and domains-local plus 3 entries in hashtags there would 5 x 5 x 3 = 75 new streams.

Usually a more targeted approach is better.

It is recommended that you put hashtags in your domains-federated or domains-local files.

Your humble author's federated file currently looks like this:

mastodon.social infosec hacker hackers osint hive lockbit hackgroup apt vicesociety

mastodon.social blackmastodon blackfediverse poc actuallyautistic neurodivergent blacklivesmatter freechina  antiracist neurodiversity  blackhistory bipoc aapi asian asianamerican pacificislander indigenous native

mastodon.social fediblock fediverse mastodev mastoadmin
mastodon.social apple politics vegan trailrunning church churchillfellowship christianity christiannationalism

My domains-local file is:

## Fake Firehose will only take local posts from these domains

mastodon.social
universeodon.com

## International English (if you aren't from the US) ###
## mastodon.scot
aus.social
mastodon.nz
respublicae.eu
mastodon.au

### Tech ###
partyon.xyz
infosec.exchange
ioc.exchange
tech.lgbt
techhub.social
fosstodon.org
appdot.net
social.linux.pizza

journa.host
climatejustice.social

This generates an acceptable stream of posts for my federated timeline. The tags I follow on mastodon.social are those that are either few in number overall, or are harder to find on local timelines.

.env.production

tl;dir, This file is fairly well commented internally, just go at it.

The sample file probably does not need any changes beyond your fakerelay information

options

fakeRelayKey

This needs to have the key you generated with fakerelay.

Example: fakeRelayKey="MrNtYH+GjwDtJtR6YCx2O4dfasdf2349QtZaVni0rsbDryETCx9lHSZmzcOAv3Y8+4LiD8bFUZbnyl4w=="

fakeRelayHost

The full URL to your fakerelay

Example: fakeRelayHost="https://fr-relay-post.myinstance.social/index"

runFirehose

This controls whether the posts will actually be sent to your relay, or only collected in your /data folder. You almost certainly want this set at:

runFirehose=true

The only reason to set it to false is for debugging, or logging posts from the fediverse.

maxCurls and minURIs

These two options are closely related. maxCurls is the maximum number of curl processes you want to have running on your system at once. If you follow timelines with a lot of posts, you may need to limit this.

Note This always needs to be higher that the total number of instances + hashtags you have configured, because each one of those is a separate curl process

fake-firehose batches posts to de-duplicate them, minURIs is the size of that batch. If you have a lot of federated posts coming in you will want to set this to a high number because a lot of them will be duplicates.

If you only use local timelines it doesn't matter, you will not have any duplicates.

It is a tradeoff between resources (and curl processes running) and how quickly you want to fill your instance's federated timeline.

Example for a moderate number of incoming posts:

## Max curl processes have not gotten out of control so this is absurdely high.
maxCurls=2000

## Nearly all of the timelines I follow are local, so there are very few duplicates.
minURIs=10

archive

Archive mode will save the json stream but not parse it, not even into URIs. This will greatly save resources, but obviously will not send it to the relay.

The only reasons to use this is for debugging or logging posts from servers.

You almost certainly want this set at:

archive=false

restartTimeout

This is how long the docker image will run before exiting. As long as your docker-compose has restart: always set this simply restarts the image to kill any hung curl processes.

The only reason to set it high is if you have a lot of timelines you follow. Each one takes time to open up, so if you restart often you will miss more posts.

Example:

restartTimeout=4h

streamDelay

This is only for debugging.

Keep it at:

streamDelay="0.1s"

Data Directory

Data is saved in the format of:

"%Y%m%d".uris.txt

In archive mode the format is:

"/data/"%Y%m%d"/"%Y%m%d".$host.json"

For example, if you set archive=true and had mastodon.social in your domains-federated or domains-local config, on January 1st, 2023 the json stream would be saved at

/data/20230101.mastodon.social.json

Misc

Backoff

An exponential backoff starts if curl fails. It is rudimentary and maxes out at 15 minutes.

DNS lookup

Before a URL starts streaming fakefirehose will look up the DNS entry of the host. If it fails, the stream will not begin, and will not attempt to begin again until the container is restarted.

Permissions

The permissions of the outputted data files will be set to root by default. This will get fixed in a future release.

Why fake firehose?

When I wrote this there were not other options I was aware of to fill a federated timeline of a small instance. The work of Gervasio Marchand is fantastic but still required programming knowledge to make use of.

I wanted the simplest setup and config I could create, without setting up an entirely new web UI.

There are a lot of things to do better, I'll work on the ones I have time and capability for. Otherwise, this project is practically begging to be re-written in python or something else.