PQX stands for Priority Queue Execution. Inspired by the official tutorial, PQX-APP uses RabbitMQ as the message system, and serves as a RPC client which pulls messages from MQ, deserializes messages and executes commands. PQX-APP can also be placed in different machines in order to execute machine-specified commands (by mailing_to field, see below).

Retry functionality is based on RabbitMQ plugin delayed_message_exchange, check this for more detail.

Bin files provided, currently:

• inspector: inspecting database table schemas, MQ settings/status and etc.

• initiator: initializing data, such as database tables, MQ settings and etc.

• rectifier: modifying initialized data (delete/recreate/...)

• subscriber: consuming message from the MQ, and execute commands

• publisher: sending message to the MQ

A full command in Json expression looks like this 🧐:

{
    "mailing_to": [
        {
            "unique_key": "h1"
        },
        {
            "unique_key": "h2",
            "common_key": "dev"
        }
    ],
    "config": {
        "retry": 5,
        "poke": 60,
        "waiting_timeout": 180,
        "consuming_timeout": 270
    },
    "cmd": {
        "CondaPython": {
            "env": "py310",
            "dir": "$HOME/Code/pqx/scripts",
            "script": "print_csv_in_line.py"
        }
    }
}

where:

• mailing_to a list of matching criteria (logic 'or', meaning this message will be sent multiple times), mailing to the queues' who match one of these criteria. if mailing_to is empty, then send to all queues (header-exchange mechanism);

• retry the number of retries, default 0;

• poke retry frequency in seconds;

• waiting_timeout the message lives in the queue (seconds), default infinity;

• consuming_timeout the acking timeout in a consumer (seconds);

• cmd the command needs to be executed, for more detail see CmdArg in adt.rs.

pub struct Command {
    pub mailing_to: Vec<HashMap<String, String>>,
    pub config: Config,
    pub cmd: CmdArg,
}

pub struct Config {
    pub retry: Option<u8>,
    pub poke: Option<u16>,
    pub waiting_timeout: Option<u32>,
    pub consuming_timeout: Option<u32>,
}

pub enum CmdArg {
    Ping {
        addr: String,
    },
    Bash {
        cmd: Vec<String>,
    },
    Ssh {
        ip: String,
        user: String,
        cmd: Vec<String>,
    },
    Sshpass {
        ip: String,
        user: String,
        pass: String,
        cmd: Vec<String>,
    },
    CondaPython {
        env: String,
        dir: String,
        script: String,
    },
    DockerExec {
        container: String,
        cmd: Vec<String>,
    },
}

• pqx-util: utilities

  • cfg: config and misc

  • db: persistent connection

  • logging: logging utils

  • mq: RabbitMQ management APIs

• pqx: library

  • ec: commands and executors

  • mq: publisher and subscriber

• pqx-app: applications

  • initiator: check existences | create tables | declare exchanges, queues and etc.

  • subscriber: app

.
├── pqx
│   └── src
│       ├── ec
│       │   ├── cmd.rs
│       │   ├── exec.rs
│       │   ├── mod.rs
│       │   └── util.rs
│       ├── mq
│       │   ├── client.rs
│       │   ├── consumer.rs
│       │   ├── mod.rs
│       │   ├── predefined.rs
│       │   ├── publish.rs
│       │   └── subscribe.rs
│       ├── error.rs
│       └── lib.rs
├── pqx-app
│   └── src
│       ├── bin
│       │   ├── initiator.rs
│       │   ├── inspector.rs
│       │   ├── publisher.rs
│       │   ├── rectifier.rs
│       │   └── subscriber.rs
│       ├── entities
│       │   ├── message_history.rs
│       │   ├── message_result.rs
│       │   └── mod.rs
│       ├── adt.rs
│       ├── cfg.rs
│       ├── exec.rs
│       ├── lib.rs
│       └── persist.rs
├── pqx-util
│   └── src
│       ├── db.rs
│       ├── error.rs
│       ├── lib.rs
│       ├── log.rs
│       ├── misc.rs
│       └── mq.rs
└── LICENSE

1. Build image for RabbitMQ (including plugins): make facilities-build

2. Make sure RabbitMQ and PostgreSQL has been started, simply by executing make facilities-setup. Check docker-compose for composing detail.

3. Add RabbitMQ user: make mq-adduser; for supervisor role (enable website operation): make mq-supervisor

4. Running the test cases

1. Following the same steps described in Tests

2. Build image for Pqx: make pqx-build

3. Create config files: make init-config, modify these configs conn.yml & init.yml.

4. Build and run a Pqx container: make pqx-build then make pqx-setup

5. Check container & initialization's availability: docker exec pqx-dev inspector -o insp

6. Create tables for message persistence and declare exchanges, queues and bindings: docker exec pqx-dev initiator -o init

7. Subscribe to a specific queue: docker exec pqx-dev ./run.sh sub start, make sure ./docker/server/config/secret.env has been filled

8. Stop a subscriber: docker exec pqx-dev ./run.sh sub stop

• cmd: cmd module, commands composition and execution

• mq: mq module, basic pub/sub

• subscriber: pub/sub combined with a command executor

• dlx: dead letter exchange

• topics: topic exchange

• headers: header exchange

• custom consumer: a further test case from subscriber, with custom consumer, command execution and logging. Moreover, a Python script for message publishing is also provided.

• callback registration: connection & channel callback registration

• delay retry: based on plugin delayed_message_exchange, implementation of message retry

• message persistence: database interaction

• mq api: RabbitMQ management APIs

• If retry happened, message would send back to the original exchange, and at the moment as a header-typed exchange, delayed-exchange would deliver this retry message to all the matched queues, which means if one consumer failed to process this message, all the other consumers would receive this message again. Hence, a strict publish header should be introduced into the message's header so that delayed-exchange could deliver the retry message to the right place.

• Delayed exchange cannot be removed unless used 'disable plugin' technique, see Makefile mq-disable-delayed-exchange, and mq-enable-delayed-exchange.

• list all consumers (simply by MqQuery)

• flexible publisher (not only read task from Json file)

• enhance Command, for instance accepting string replacement in CmdArg

• Module dynamic: dynamically set/del exchange/queue/binding

• message aggregator, see this