A Solana indexer providing fast, accurate account information

As a message producer, the Holaplex Indexing Service leverages the geyser-plugin-interface to send accounts data directly to a RabbitMQ instance. As a message consumer, the indexer consumes these account messages, deserializes them and inserts them into a PostgreSQL database. Each account needs its own processor, schema and model.

This dataset is derived entirely from the messages produced by a validator. This supports a unidirectional dataflow. All data goes directly to the Solana blockchain before it is saved in any off chain storage.

• Solana Geyser plugin, responsible for sending data to our queue system
• RabbitMQ consumers, responsible for parsing messages and routing them to the proper processor
• PostgreSQL database, saves the deserialized data
• GraphQL Crate - serves the PostgreSQL data

Currently, the indexer covers the following Solana programs:

• Holaplex wallet graph program
• Metaplex program
• Metaplex auction program
• Metaplex auction house program
• Metaplex candy machine program
• Metaplex metadata program
• SPL token program

Additionally, the following off-chain data is also indexed:

• Holaplex storefronts
• Holaplex marketplaces
• Metaplex JSON metadata

To set up a development environment, you will need rustup, Cargo, Docker, docker-compose, and the Diesel CLI. Specifically, you will need diesel_cli installed with the postgres feature, which can be done like so:

$ cargo install diesel_cli --version 1.4.1 --no-default-features --features postgres

Installing diesel will require libpq to be on your system (brew install postgresql on Mac).

Once you have the required dependencies, you can get started by running the following script to initialize and migrate a containerized Postgres database in the background:

$ ./start-developing.sh

All indexer crates attempt to connect to the database by reading a Postgres URI from one of three environment variables:

• DATABASE_READ_URL is used by the GraphQL server to identify a read-only database.
• DATABASE_WRITE_URL is used by all indexer services to identify a writable database.
• DATABASE_URL is used as a fallback by all crates, and is assumed (but not guaranteed) to be writeable.

For debug builds the .env* files provided in the repository will provide a default connection string pointed at the database defined in docker-compose.yml. For production builds the database must be manually configured according to the environment variables above.

The indexer consists of four services run by two binaries and a Geyser plugin. All services are connected via a common RabbitMQ node.

To build the Geyser plugin,clone this repo https://github.com/holaplex/indexer-geyser-plugin.git and use the following build command:

$ cargo build -pholaplex-indexer-rabbitmq-geyser

This will produce a build artifact named libholaplex-indexer-rabbitmq-geyser with the appropriate file extension for a dynamic library for the host system (i.e. .dll, .dylib, or .so). This plugin can then be used with a Solana validator. A sample Geyser JSON configuration for the plugin can be found in crates/geyser-rabbitmq/sample_config.json.

Once the plugin is up and running, the three indexer consumer services can be launched to process messages from the validator. The consumers can be launched as follows:

$ cargo run --bin holaplex-indexer-geyser --features geyser &
$ cargo run --bin holaplex-indexer-http --features http -- --entity metadata-json &
$ cargo run --bin holaplex-indexer-http --features http -- --entity store-config &

All services will need to be configured to run with the same settings that the Geyser plugin was configured with, otherwise they will receive no messages or simply fail to start.

i.e : if your geyser config json has "network": "mainnet" and "startup": null, then the exchange name will be mainnet.startup-all.accounts and to connect to it you'll need to pass --network mainnet --startup all to the geyser-consumer binary (or put NETWORK=mainnet and STARTUP=all in .env.local)

The server binds to the address [::]:3000 by default. To change this, set the -p argument/PORT environment variable or the --addr argument/ADDRESS environment variable.

To see more options for the server, run the following:

$ cargo run --bin holaplex-indexer-graphql -- --help

To launch the GraphQL server, simply run the following:

$ cargo run --bin holaplex-indexer-graphql

Before pushing branch changes, run the following (or add it to your Git pre-push hook) to check for problems, style errors, and schema issues:

$ scripts/pre-push.sh