A program designed to fix the disk-space consumption issues that matrix-synapse continues to suffer
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
forest 5ca631067a add goroutine names to panic-safe wrapper 5 months ago
deployment bugfix 2 years ago
frontend add actual chartjs build 2 years ago
readme flesh out the readme 2 years ago
.gitignore ignore config.json 2 years ago
ReadMe.md flesh out the readme 2 years ago
config.example.json ignore config.json 2 years ago
db_model.go add goroutine names to panic-safe wrapper 5 months ago
disk_space.go fix issue where delete room fails due to timeout / retry fails due to 10 months ago
frontend.go add goroutine names to panic-safe wrapper 5 months ago
go.mod copypasting frontend from greenhouse + working on login 2 years ago
go.sum copypasting frontend from greenhouse + working on login 2 years ago
main.go add goroutine names to panic-safe wrapper 5 months ago
matrix_admin_service.go remove spurious room 404 error messages 2 years ago
storage_service.go add deployment code 2 years ago

ReadMe.md

matrix-synapse-diskspace-janitor

scruffy the janitor from futurama

toilets and boilers, boilers and toilets

screenshot showing three pie charts, disk space,  disk usage by table, and disk usage by room. A form with a list of rooms  having DELETE and BAN checkboxes sits below the charts.


We did come up with some shell scripts to handle this, but it was an annoying recurring manual maintenance burden. Due to the complicated, multi-step nature of the cleanup process, I ended up creating an application to handle it instead of continuing to try to script it. Yes, its probably overkill, but it was fun. And who knows, maybe it can be useful to someone else.


Configuration Overview

config.json

{
  "FrontendPort": 6712,
  "FrontendDomain": "matrix-diskspace-janitor.cyberia.club",
  "MatrixServerPublicDomain": "cyberia.club",
  "MatrixURL": "http://localhost:8008",
  "AdminMatrixRoomId": "!oAVmChLLsrnfaSubLP:cyberia.club",
  "MatrixAdminToken": "syt_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
  "DatabaseType": "postgres",

  "DatabaseConnectionString": 
  "host=localhost port=5432 user=synapse_user password=xxxxxxxxxx database=synapse sslmode=disable",

  "MediaFolder": "/var/lib/matrix-synapse",
  "PostgresFolder": "/var/lib/postgresql"
}

FrontendDomain

This is the domain under which the janitor web application will be served.

MatrixServerPublicDomain

This is the domain part of matrix IDs on your homeserver. For example, my matrix ID is @forestjohnson:cyberia.club so I use cyberia.club as my MatrixServerPublicDomain.

AdminMatrixRoomId

Users who are in the AdminMatrixRoomId private room will be able to log into the tool with their matrix account. (note, the account has to be on the same homeserver)


It uses https://git.sequentialread.com/forest/config-lite for its configuration, so you can also use environment variables to configure it, for example, when you are using docker:

  blog-external-service2:
    image: matrix-synapse-diskspace-janitor:test
    restart: always
    environment:
      JANITOR_FRONTENDPORT: 6712
      JANITOR_FRONTENDDOMAIN: 'matrix-diskspace-janitor.cyberia.club'
      ... etc

Origin Story

Matrix-synapse (the matrix homeserver implementation) requires a postgres database server to operate. It stores a lot of stuff in this postgres database, information about all the rooms that users on the server have joined, etc.

The problem at hand:

Matrix-synapse stores a lot of data that it has no way of cleaning up or deleting.

Specifically, there is a table it creates in the database called state_groups_state:

root@matrix:~# sudo -u postgres pg_dump synapse -t state_groups_state --schema-only
--
-- PostgreSQL database dump
--
...

CREATE TABLE public.state_groups_state (
    state_group bigint NOT NULL,
    room_id text NOT NULL,
    type text NOT NULL,
    state_key text NOT NULL,
    event_id text NOT NULL
);

I don't understand what this table is for, however, I can recognize fairly easily that it accounts for the grand majority of the disk space bloat of a matrix-synapse instance:

top 10 tables by disk space used, cyberia.club instance:

a pie chart showing state_groups_state using 87% of the disk space

So, I think it's safe to say that if we can cut down the size of state_groups_state, then we can solve our disk space issues.

I know that there are other projects dedicated to this, like https://github.com/matrix-org/rust-synapse-compress-state

However, a cursory examination of the data in state_groups_state led me to believe maybe there is an easier and better way.

state_groups_state DOES have a room_id column on it. It's not indexed by room_id, but we can still count the # of rows for each room and rank them:

top 100 rooms by number of state_groups_state rows, cyberia.club instance:

a pie chart with two slices taking up about 2 thirds of the pie, and the remaining third taken up mostly by the next 8 slices

In summary, it looks like

about 90% of the disk space used by matrix-synapse is in state_groups_state, and about 90% of the rows in state_groups_state come from just a handfull of rooms.

So from this information we have hatched a plan:

Just delete those rooms from our homeserver 4head

However, unfortunately the matrix-synapse delete room API does not remove anything from state_groups_state.

This is similar to the way that the matrix-synapse message retention policies also do not remove anything from state_groups_state.

In fact, probably helps explain why state_groups_state gets hundreds of millions of rows and takes up so much disk space: Nothing ever deletes from it!!

We did come up with some shell scripts to handle this, but it was an annoying recurring manual maintenance burden. Due to the complicated, multi-step nature of the cleanup process, I ended up creating an application to handle it instead of continuing to try to script it. Yes, its probably overkill, but it was fun. And who knows, maybe it can be useful to someone else.