This repository provides a Docker container for the rust-synapse-compress-state project developed by Matrix.org. The primary purpose is to offer a convenient way to run the state compression tool for Synapse without dealing with manual installations and dependencies.

• GitHub Repository
• Docker Hub Repository

PostgreSQL with TCP port:

docker pull tcpipuk/rust-synapse-compress-state:latest
docker run -e POSTGRES_USER="synapse" -e POSTGRES_PASSWORD="YOUR_PASSWORD" -e POSTGRES_DB="synapse" -e POSTGRES_HOST="db" -e POSTGRES_PORT="5432" tcpipuk/rust-synapse-compress-state:latest

PostgreSQL with Unix sockets:

docker pull tcpipuk/rust-synapse-compress-state:latest
docker run -e POSTGRES_USER="synapse" -e POSTGRES_PASSWORD="YOUR_PASSWORD" -e POSTGRES_DB="db" -e POSTGRES_PATH="/path/to/socket/dir" tcpipuk/rust-synapse-compress-state:latest

Example docker-compose.yml for PostgreSQL with TCP port:

version: '3'

services:
  synapse-compress-state:
    image: tcpipuk/rust-synapse-compress-state:latest
    environment:
      POSTGRES_USER: synapse
      POSTGRES_PASSWORD: YOUR_PASSWORD
      POSTGRES_DB: synapse
      POSTGRES_HOST: db
      POSTGRES_PORT: 5432

Example docker-compose.yml for PostgreSQL with Unix sockets:

version: '3'

services:
  synapse-compress-state:
    image: tcpipuk/rust-synapse-compress-state:latest
    environment:
      POSTGRES_USER: synapse
      POSTGRES_PASSWORD: YOUR_PASSWORD
      POSTGRES_DB: synapse
      POSTGRES_PATH: /path/to/socket/dir

• CHUNK_SIZE: Determines the number of state groups to process at once. This is particularly useful for machines with limited memory as smaller chunk sizes can be set. If no space savings are found for the entire chunk, it's skipped. Default: "500".

• CHUNKS_TO_COMPRESS: Specifies the number of chunks (of size CHUNK_SIZE) to compress. The higher this value, the longer the compression process will run. Default: "100".

• COMPRESSION_LEVELS: Defines the sizes of each new level in the compression algorithm as a comma-separated list. The list's first entry is for the most granular level, with each subsequent entry for the next highest level. The total number of entries determines the algorithm's levels. The sum of the sizes impacts the state fetching performance from the database, as it sets the upper limit on the iterations needed to fetch a specific state set. Default: "100,50,25".

The tool may exit with specific codes, indicating various states or issues:

• Exit Code 0: Normal exit. The compressor has completed its task without any errors.
• Exit Code 101: Indicates a Rust panic, often due to the room it was previously compressing being purged from the Synapse database. If RECOVER_AUTOMATICALLY=1 is set, the tool will attempt to automatically recover by dropping the state_compressor_progress, state_compressor_state, and state_compressor_total_progress tables.
• Exit Code 137: This typically indicates that the compressor was terminated by the system, likely due to reaching memory limits. The defaults can consume over 1GB of memory on a large room, so consider checking your Docker configured limits and possibly lowering the CHUNK_SIZE or CHUNKS_TO_COMPRESS values.

Setting RECOVER_AUTOMATICALLY=1 in environment variables enables the tool to automatically drop the state_compressor_progress, state_compressor_state, and state_compressor_total_progress tables and start from scratch when critical failures (error code 101) occur. While this can resolve common issues, such as a room being purged from the Synapse database, it should be used with caution as it involves modifying the database schema - users enable this feature at their own risk.