@ovotech/castle
A kafka and avro based event listener
Last updated a month ago by ikerin .
Apache-2.0 · Repository · Bugs · Original npm · Tarball · package.json
$ cnpm install @ovotech/castle 
SYNC missed versions from official npm registry.

Castle

A framework around Kafka.js to transparently use Schema Registry and create an application that consumes, produces, and reacts to different kafka topics. Supports consumption in batches or in parallel. Statically define and verify the schemas / message types in TypeScript

Usage

yarn add @ovotech/castle

examples/simple.ts

import { createCastle, produce, consumeEachMessage, describeCastle } from '@ovotech/castle';
import { Event, EventSchema } from './avro';

// Define producers as pure functions
// With statically setting the typescript types and avro schemas
const mySender = produce<Event>({ topic: 'my-topic-1', schema: EventSchema });

// Define consumers as pure functions
// With statically setting which types it will accept
const eachEvent = consumeEachMessage<Event>(async ({ message }) => {
  console.log(message.value);
});

const main = async () => {
  const castle = createCastle({
    schemaRegistry: { uri: 'http://localhost:8081' },
    kafka: { brokers: ['localhost:29092'] },
    consumers: [{ topic: 'my-topic-1', groupId: 'my-group-1', eachMessage: eachEvent }],
  });

  // Start all consumers and producers
  await castle.start();

  console.log(describeCastle(castle));

  await mySender(castle.producer, [{ value: { field1: 'my-string' } }]);
};

main();

You can connect to multiple topics, each of which is will have its own independent consumer group.

You can also use schema registry directly to encode and decode messages.

examples/multiple-topics.ts

import {
  createCastle,
  produce,
  consumeEachMessage,
  consumeEachBatch,
  describeCastle,
} from '@ovotech/castle';
import {
  StartEvent,
  StartEventSchema,
  FeedbackEvent,
  FeedbackEventSchema,
  CompleteEvent,
  CompleteEventSchema,
} from './avro';

enum Topic {
  Start = 'start',
  Complete = 'complete',
  Feedback = 'feedback',
  Batched = 'batched',
}

// Define multiple producers as pure functions
const sendStart = produce<StartEvent>({ topic: Topic.Start, schema: StartEventSchema });
const sendComplete = produce<CompleteEvent>({ topic: Topic.Complete, schema: CompleteEventSchema });
const sendFeedback = produce<FeedbackEvent>({ topic: Topic.Feedback, schema: FeedbackEventSchema });

// Define a consumer as a pure function
const eachStartEvent = consumeEachMessage<StartEvent>(async ({ message }) => {
  console.log(`Started Processing ${message.value.id}`);
});

// Define a batch consumer as a pure function
const eachBatchFeedbackEvent = consumeEachBatch<FeedbackEvent>(async ({ batch, producer }) => {
  console.log(`Feedback ${batch.messages.map(msg => `${msg.value.id}:${msg.value.status}`)}`);
  console.log('Sending complete events');
  sendComplete(
    producer,
    batch.messages.map(msg => ({ value: { id: msg.value.id } })),
  );
});

// Define a parallel consumer as a pure function
const eachCompleteEvent = consumeEachMessage<CompleteEvent>(async ({ message }) => {
  console.log(`Completed ${message.value.id}`);
});

const eachSizedBatch = consumeEachBatch(async ({ batch: { messages, partition } }) =>
  console.log('Batch Size', partition, messages.length),
);

const main = async () => {
  const castle = createCastle({
    // Setup topic aliases
    // You can use short statically checked names in the code,
    // but configure long environment specific kafka topic names
    topicsAlias: {
      [Topic.Start]: 'start-topic-name-1',
      [Topic.Feedback]: 'feedback-topic-name-1',
      [Topic.Complete]: 'complete-topic-name-1',
    },
    schemaRegistry: { uri: 'http://localhost:8081' },
    kafka: { brokers: ['localhost:29092'] },
    consumers: [
      {
        topic: Topic.Start,
        groupId: 'start-group-1',
        eachMessage: eachStartEvent,
      },
      {
        topic: Topic.Feedback,
        groupId: 'feedback-group-1',
        eachBatch: eachBatchFeedbackEvent,
      },
      {
        topic: Topic.Complete,
        groupId: 'complete-group-1',
        partitionsConsumedConcurrently: 2,
        eachMessage: eachCompleteEvent,
      },
      {
        topic: Topic.Batched,
        groupId: 'batched-group-1',
        /* Use eachSizedBatch to instruct castle to break down
         * Kafkajs batches into chunks of size up to maxBatchSize.
         *
         * Castle will handle heartbeats and nudging kafka to commitIfNecessary
         * after each chunk is processed.
         *
         * The consumer will be called sequentially for each chunk within
         * a given partition.
         */
        eachSizedBatch,
        maxBatchSize: 50,
      },
    ],
  });

  await castle.start();

  console.log(describeCastle(castle));

  // Perform a siqeunce of events
  // - send start events, wait a bit
  await sendStart(castle.producer, [{ value: { id: 10 } }, { value: { id: 20 } }]);

  // - wait a bit
  await new Promise(resolve => setTimeout(resolve, 1000));

  // - send feedback events which would produce the complete events
  await sendFeedback(castle.producer, [
    { value: { id: 10, status: 'Sent' } },
    { value: { id: 20, status: 'Failed' } },
  ]);
};

main();

Middlewares

Castle is designed to help building complex applications, where you want to share logic between consumers. This is achieved with building middlewares that process each consumer and add / modify it. Statically verified by typescript

examples/with-middlewares.ts

import { createCastle, describeCastle, produce, consumeEachMessage } from '@ovotech/castle';
import { StartEvent, StartEventSchema, CompleteEvent, CompleteEventSchema } from './avro';
import {
  createDb,
  createLogging,
  DbContext,
  LoggingContext,
  createErrorHandling,
} from './middlewares';

const start = produce<StartEvent>({ topic: 'my-start-3', schema: StartEventSchema });
const complete = produce<CompleteEvent>({ topic: 'my-complete-3', schema: CompleteEventSchema });

const eachStart = consumeEachMessage<StartEvent, DbContext & LoggingContext>(
  async ({ message, db, logger, producer }) => {
    logger.log('Started', message.value.id);
    const { rows } = await db.query('SELECT avatar FROM users WHERE id = $1', [message.value.id]);
    logger.log('Found', rows, 'Sending Complete');
    complete(producer, [{ value: { id: message.value.id } }]);
  },
);

const eachComplete = consumeEachMessage<CompleteEvent, LoggingContext>(
  async ({ message, logger }) => {
    logger.log('Complete received for', message.value.id);
  },
);

const main = async () => {
  const db = createDb({
    user: 'boost-statements-api',
    database: 'boost-statements-api',
    password: 'dev-pass',
    host: '127.0.0.1',
  });
  const logging = createLogging(console);
  const errorHandling = createErrorHandling();

  const castle = createCastle({
    schemaRegistry: { uri: 'http://localhost:8081' },
    kafka: { brokers: ['localhost:29092'] },
    consumers: [
      {
        topic: 'my-start-3',
        groupId: 'my-start-3',
        eachMessage: logging(errorHandling(db(eachStart))),
      },
      {
        topic: 'my-complete-3',
        groupId: 'my-complete-3',
        eachMessage: logging(errorHandling(eachComplete)),
      },
    ],
  });

  await castle.start();

  console.log(describeCastle(castle));

  await start(castle.producer, [{ value: { id: 1 } }]);
};

main();

Optional consumption

You can consume the topics optionally with the topic optionalConsumers function

examples/optional.ts

import {
  createCastle,
  produce,
  consumeEachMessage,
  describeCastle,
  optionalConsumers,
} from '@ovotech/castle';
import { Event, EventSchema } from './avro';

// Allow topic name for consumer to be undefined. If it ever is undefined, no consumption will happen.
const consumeTopic: string | undefined = 'my-topic-1';

// Define producers as pure functions
// With statically setting the typescript types and avro schemas
const mySender = produce<Event>({ topic: 'my-topic-1', schema: EventSchema });

// Define consumers as pure functions
// With statically setting which types it will accept
const eachEvent = consumeEachMessage<Event>(async ({ message }) => {
  console.log(message.value);
});

const main = async () => {
  const castle = createCastle({
    schemaRegistry: { uri: 'http://localhost:8081' },
    kafka: { brokers: ['localhost:29092'] },
    consumers: optionalConsumers([
      { topic: consumeTopic, groupId: 'my-group-1', eachMessage: eachEvent },
    ]),
  });

  // Start all consumers and producers
  await castle.start();

  console.log(describeCastle(castle));

  await mySender(castle.producer, [{ value: { field1: 'my-string' } }]);
};

main();

Splitting the castle

Sometimes you would want to use the producer / consumers instances independantly, before you assemble the main castle instance.

For example if you also have a service that uses castle to produce messages, but the castle instance needs to use that service too. You can split the castle instance, creating the producer first, passing it to where its needed, then combining the rest into a castle sintance.

examples/parts.ts

import {
  produce,
  consumeEachMessage,
  describeCastle,
  createKafka,
  createProducer,
  createConsumers,
  createCastleFromParts,
} from '@ovotech/castle';
import { Event, EventSchema } from './avro';

// Define producers as pure functions
// With statically setting the typescript types and avro schemas
const mySender = produce<Event>({ topic: 'my-topic-1', schema: EventSchema });

// Define consumers as pure functions
// With statically setting which types it will accept
const eachEvent = consumeEachMessage<Event>(async ({ message }) => {
  console.log(message.value);
});

const main = async () => {
  const kafka = createKafka({
    schemaRegistry: { uri: 'http://localhost:8081' },
    kafka: { brokers: ['localhost:29092'] },
  });
  const producer = createProducer(kafka);
  const consumers = createConsumers(kafka, [
    { topic: 'my-topic-1', groupId: 'my-group-1', eachMessage: eachEvent },
  ]);
  const castle = createCastleFromParts({ kafka, producer, consumers });

  // Start all consumers and producers
  await castle.start();

  console.log(describeCastle(castle));

  // You can use the stand alone producer elsewhere
  await mySender(producer, [{ value: { field1: 'my-string' } }]);
};

main();

Running the tests

You can run the tests with:

yarn test

Coding style (linting, etc) tests

Style is maintained with prettier and eslint

yarn lint

Deployment

Deployment is preferment by lerna automatically on merge / push to master, but you'll need to bump the package version numbers yourself. Only updated packages with newer versions will be pushed to the npm registry.

Contributing

Have a bug? File an issue with a simple example that reproduces this so we can take a look & confirm.

Want to make a change? Submit a PR, explain why it's useful, and make sure you've updated the docs (this file) and the tests (see test folder).

License

This project is licensed under Apache 2 - see the LICENSE file for details

Current Tags

  • 0.6.2                                ...           latest (a month ago)

24 Versions

  • 0.6.2                                ...           a month ago
  • 0.6.1                                ...           2 months ago
  • 0.6.0                                ...           2 months ago
  • 0.5.2                                ...           2 months ago
  • 0.5.1                                ...           4 months ago
  • 0.5.0                                ...           4 months ago
  • 0.4.7                                ...           5 months ago
  • 0.4.6                                ...           6 months ago
  • 0.4.5                                ...           6 months ago
  • 0.4.4                                ...           7 months ago
  • 0.4.3                                ...           7 months ago
  • 0.4.2                                ...           8 months ago
  • 0.4.1                                ...           8 months ago
  • 0.4.0                                ...           8 months ago
  • 0.3.1                                ...           9 months ago
  • 0.3.0                                ...           10 months ago
  • 0.2.2                                ...           10 months ago
  • 0.2.1                                ...           10 months ago
  • 0.2.0                                ...           10 months ago
  • 0.1.4                                ...           10 months ago
  • 0.1.3                                ...           a year ago
  • 0.1.2                                ...           a year ago
  • 0.1.1                                ...           a year ago
  • 0.1.0                                ...           a year ago
Downloads
Today 1
This Week 1
This Month 1
Last Day 0
Last Week 0
Last Month 11
Dependencies (3)
Dev Dependencies (19)
Dependents (1)

Copyright 2014 - 2017 © taobao.org |