Skip to content

@hyperledger/cactus-cmd-api-server

Summary

This package is part of the Hyperledger Cactus blockchain integration framework and is used as a shell/container of sort for housing different Cactus plugins (which all live in their own npm packages as well).

The API server gives you for free the following benefits, should you choose to use it:

  1. Automatic wiring of API endpoints for Cactus plugins which implement the IPluginWebService Typescript interface

  2. Lightweight inversion of control container provided to plugins in the form of the PluginRegistry so that plugins can depend on each other in a way that each plugin instance can be uniquely identified and obtained by other plugins. A great example of this in action is ledger connector plugins frequently using the PluginRegistry to look up instances of keychain plugins to get access to secrets that are needed for the connector plugins to accomplish certain tasks such as cryptographically signing some information or SSH-ing into a server instance in order to upload and deploy binary (or otherwise) artifacts of smart contracts.

Usage

Like with most parts of the framework in Cactus, using the ApiServer is optional.

To see the ApiServer in action, the end to end tests of the framework are a great place to start. A few excerpts that regularly occur in said tests can be seen below as well for the reader’s convenience.

One of our design principles for the framework is secure by default which means that the API servers

  1. assumes TLS is enabled by default and

  2. cross-origin resource sharing is disabled completely

Basic Example

#!/usr/bin/env node

import { ApiServer } from "../api-server";
import { ConfigService } from "../config/config-service";
import { Logger, LoggerProvider } from "@hyperledger/cactus-common";

const log: Logger \= LoggerProvider.getOrCreate({
  label: "cactus-api",
  level: "INFO",
});

const main \= async () \=> {
  const configService \= new ConfigService();
  const config \= await configService.getOrCreate();
  const serverOptions \= config.getProperties();

  LoggerProvider.setLogLevel(serverOptions.logLevel);

  if (process.argv\[2\].includes("help")) {
    const helpText \= ConfigService.getHelpText();
    console.log(helpText);
    log.info(\`Effective Configuration:\`);
    log.info(JSON.stringify(serverOptions, null, 4));
  } else {
    const apiServer \= new ApiServer({ config: serverOptions });
    await apiServer.start();
  }
};

export async function launchApp(): Promise<void\> {
  try {
    await main();
    log.info(\`Cactus API server launched OK \`);
  } catch (ex) {
    log.error(\`Cactus API server crashed: \`, ex);
    process.exit(1);
  }
}

if (require.main \=== module) {
  launchApp();
}

Remote Plugin Imports at Runtime Example

import { PluginImportType, PluginImportAction } from "@hyperledger/cactus-core-api";
import { ApiServer } from "@hyperledger/cactus-cmd-api-server";
import { ConfigService } from "@hyperledger/cactus-cmd-api-server";
import { Logger, LoggerProvider } from "@hyperledger/cactus-common";

const main \= async () \=> {

  const configService \= new ConfigService();
  const apiServerOptions \= await configService.newExampleConfig();
  // If there is no configuration file on the file system, just set it to empty string
  apiServerOptions.configFile \= "";
  // Enable CORS for
  apiServerOptions.apiCorsDomainCsv \= "your.domain.example.com";
  apiServerOptions.apiPort \= 3000;
  apiServerOptions.cockpitPort \= 3100;
  apiServerOptions.grpcPort \= 5000;
  // Disble TLS (or provide TLS certs for secure HTTP if you are deploying to production)
  apiServerOptions.apiTlsEnabled \= false;
  apiServerOptions.plugins \= \[
    {
      // npm package name of the plugin you are installing
      // Since this will be imported at runtime, you are responsible for
      // installing the package yourself prior to launching the API server.
      packageName: "@hyperledger/cactus-plugin-keychain-vault",
      // The REMOTE value means that a different plugin factory will be imported and
      // called to obtain the plugin instance. This way plugins can support them
      // being imported by the API server regardless of the language the plugin
      // was written in.
      type: PluginImportType.REMOTE,
      // The INSTALL value means that the plugin will be installed instead of
      // only instantiate it
      action: PluginImportAction.INSTALL,
      // The options that will be passed in to the plugin factory
      options: {
        keychainId: "\_keychainId\_",
        instanceId: "\_instanceId\_",
        remoteConfig: configuration,
      },
    },
  \];
  const config \= await configService.newExampleConfigConvict(apiServerOptions);

  const apiServer \= new ApiServer({
    config: config.getProperties(),
  });

  // start the API server here and you are ready to roll
};

export async function launchApp(): Promise<void\> {
  try {
    await main();
    log.info(\`Cactus API server launched OK \`);
  } catch (ex) {
    log.error(\`Cactus API server crashed: \`, ex);
    process.exit(1);
  }
}

if (require.main \=== module) {
  launchApp();
}

Complete Example

For a complete example of how to use the API server, read all the code of the supply chain example’s backend package:

https://github.com/hyperledger/cactus/tree/main/examples/cactus-example-supply-chain-backend/src/main/typescript

Deployment Scenarios

There’s a set of building blocks (members, nodes, API server processes, plugin instances) that you can use when defining (founding) a consortium and these building blocks relate to each other in a way that can be expressed with an entity relationship diagram which can be seen below. The composability rules can be deducted from how the diagram elements (entities) are connected (related) to each other, e.g. the API server process can have any number of plugin instances in it and a node can contain any number of API server processes, and so on until the top level construct is reached: the consortium.

Consortium management does not relate to achieving consensus on data/transactions involving individual ledgers, merely about consensus on the metadata of a consortium.

deployment-entity-relationship-diagram.png

Now, with these composability rules in mind, let us demonstrate a few different deployment scenarios (both expected and exotic ones) to showcase the framework’s flexibility in this regard.

Production Deployment Example

Many different configurations are possible here as well. One way to have two members form a consortium and both of those members provide highly available, high throughput services is to have a deployment as shown on the below figure. What is important to note here is that this consortium has 2 nodes, 1 for each member and it is irrelevant how many API servers those nodes have internally because they all respond to requests through the network host/web domain that is tied to the node. One could say that API servers do not have a distinguishable identity relative to their peer API servers, only the higher-level nodes do.

deployment-production-example.png

Low Resource Deployment Example

This is an example to showcase how you can pull up a full consortium even from within a single operating system process (API server) with multiple members and their respective nodes. It is not something that’s recommended for a production grade environment, ever, but it is great for demos and integration tests where you have to simulate a fully functioning consortium with as little hardware footprint as possible to save on time and cost.

The individual nodes/API servers are isolated by listening on separate TCP ports of the machine they are hosted on:

deployment-low-resource-example.png

Containerization

Building the container image locally

In the Cactus project root say:

DOCKER_BUILDKIT=1 docker build -f ./packages/cactus-cmd-api-server/Dockerfile . -t cas -t cactus-api-server

Build with a specific version of the npm package:

DOCKER_BUILDKIT=1 docker build --build-arg NPM_PKG_VERSION=main -f ./packages/cactus-cmd-api-server/Dockerfile . -t cas -t cactus-api-server

Running the container image locally

Before running the examples here you need to build the image locally. See section Building the container image locally for details on how to do that.

Once you’ve built the container, the following commands should work:

  • Launch container - no plugins, default configuration

    docker run \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ cas

  • Launch container with plugins of your choice (keychain, consortium connector, etc.)

    docker run \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ cas \ ./node_modules/.bin/cactusapi \ --plugins='[{"packageName": "hyperledger/cactus-plugin-ledger-connector-fabric", "type": "org.hyperledger.cactus.plugin_import_type.LOCAL", "action": "org.hyperledger.cactus.plugin_import_action.INSTALL", "options": { "connectionProfile": {}, "instanceId": "some-unique-instance-id"}}]'

  • Launch container with plugin configuration as an environment variable:

    docker run \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ --env PLUGINS='[{"packageName": "hyperledger/cactus-plugin-ledger-connector-besu", "type": "org.hyperledger.cactus.plugin_import_type.LOCAL", "action": "org.hyperledger.cactus.plugin_import_action.INSTALL", "options": {"rpcApiHttpHost": "http://localhost:8545", "instanceId": "some-unique-besu-connector-instance-id"}}]' \ cas

  • Launch container with plugin configuration as a CLI argument:

    docker run \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ cas \ ./node_modules/.bin/cactusapi \ --plugins='[{"packageName": "hyperledger/cactus-plugin-ledger-connector-besu", "type": "org.hyperledger.cactus.plugin_import_type.LOCAL", "action": "org.hyperledger.cactus.plugin_import_action.INSTALL", "options": {"rpcApiHttpHost": "http://localhost:8545", "instanceId": "some-unique-besu-connector-instance-id"}}]'

  • Launch container with configuration file mounted from host machine:

    echo '[{"packageName": "hyperledger/cactus-plugin-ledger-connector-besu", "type": "org.hyperledger.cactus.plugin_import_type.LOCAL", "action": "org.hyperledger.cactus.plugin_import_action.INSTALL", "options": {"rpcApiHttpHost": "http://localhost:8545", "instanceId": "some-unique-besu-connector-instance-id"}}]' > cactus.json

    docker run \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ --mount type=bind,source="$(pwd)"/cactus.json,target=/cactus.json \ cas \ ./node_modules/.bin/cactusapi \ --config-file=/cactus.json

Testing API calls with the container

Don’t have a Besu network on hand to test with? Test or develop against our Besu All-In-One container!

  1. Terminal Window 1 (Ledger)

    docker run --publish 8545:8545 hyperledger/cactus-besu-all-in-one:latest

  2. Terminal Window 2 (Cactus API Server)

    docker run \ --network host \ --rm \ --publish 3000:3000 \ --publish 4000:4000 \ --env AUTHORIZATION_PROTOCOL='NONE' \ --env AUTHORIZATION_CONFIG_JSON='{}' \ --env GRPC_TLS_ENABLED=false \ --env PLUGINS='[{"packageName": "hyperledger/cactus-plugin-ledger-connector-besu", "type": "org.hyperledger.cactus.plugin_import_type.LOCAL", "action": "org.hyperledger.cactus.plugin_import_action.INSTALL", "options": {"rpcApiHttpHost": "http://localhost:8545", "instanceId": "some-unique-besu-connector-instance-id"}}]' \ cas

  3. Terminal Window 3 (curl - replace eth accounts as needed)

    curl --location --request POST 'http://127.0.0.1:4000/api/v1/plugins/@hyperledger/cactus-plugin-ledger-connector-besu/run-transaction' \ --header 'Content-Type: application/json' \ --data-raw '{ "web3SigningCredential": { "ethAccount": "627306090abaB3A6e1400e9345bC60c78a8BEf57", "secret": "c87509a1c067bbde78beb793e6fa76530b6382a4c0241e5e4a9ec0a0f44dc0d3", "type": "PRIVATE_KEY_HEX" }, "consistencyStrategy": { "blockConfirmations": 0, "receiptType": "NODE_TX_POOL_ACK" }, "transactionConfig": { "from": "627306090abaB3A6e1400e9345bC60c78a8BEf57", "to": "f17f52151EbEF6C7334FAD080c5704D77216b732", "value": 1, "gas": 10000000 } }'

  4. The above should produce a response that looks similar to this:

    { "success": true, "data": { "transactionReceipt": { "blockHash": "0x7c97c038a5d3bd84613fe23ed442695276d5d2df97f4e7c4f10ca06765033ffd", "blockNumber": 1218, "contractAddress": null, "cumulativeGasUsed": 21000, "from": "0x627306090abab3a6e1400e9345bc60c78a8bef57", "gasUsed": 21000, "logs": [], "logsBloom": "0x00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", "status": true, "to": "0xf17f52151ebef6c7334fad080c5704d77216b732", "transactionHash": "0xc7fcb46c735bdc696d500bfc70c72595a2b8c31813929e5c61d9a5aec3376d6f", "transactionIndex": 0 } } }

Prometheus Exporter

This class creates a prometheus exporter, which scrapes the total Cactus node count.

Usage Prometheus

The prometheus exporter object is initialized in the ApiServer class constructor itself, so instantiating the object of the ApiServer class, gives access to the exporter object. You can also initialize the prometheus exporter object separately and then pass it to the IApiServerConstructorOptions interface for ApiServer constructor.

getPrometheusMetricsV1 function returns the prometheus exporter metrics, currently displaying the total plugins imported, which currently refreshes to match the plugin count, everytime setTotalPluginImports method is called.

Prometheus Integration

To use Prometheus with this exporter make sure to install Prometheus main component. Once Prometheus is setup, the corresponding scrape_config needs to be added to the prometheus.yml

- job_name: 'consortium_manual_exporter' metrics_path: /api/v1/api-server/get-prometheus-exporter-metrics scrape_interval: 5s static_configs: - targets: ['{host}:{port}']

Here the host:port is where the prometheus exporter metrics are exposed. The test cases (For example, packages/cactus-plugin-consortium-manual/src/test/typescript/unit/consortium/get-node-jws-endpoint-v1.test.ts) exposes it over 0.0.0.0 and a random port(). The random port can be found in the running logs of the test case and looks like (42379 in the below mentioned URL) Metrics URL: http://0.0.0.0:42379/api/v1/api-server/get-prometheus-exporter-metrics/get-prometheus-exporter-metrics

Once edited, you can start the prometheus service by referencing the above edited prometheus.yml file. On the prometheus graphical interface (defaulted to http://localhost:9090), choose Graph from the menu bar, then select the Console tab. From the Insert metric at cursor drop down, select cactus_api_server_total_plugin_imports and click execute

Shutdown Hook

The API config contains a flag:

{
  "enableShutdownHook": true
}

This allows for graceful shutdown of the API server after a SIGINT via cli CTRL + C. This hook can be disabled by passing in false either via the TypeScript constructor or the JSON config file.

Helper code

response.type.ts

This file contains the various responses of the metrics.

data-fetcher.ts

This file contains functions encasing the logic to process the data points.

metrics.ts

This file lists all the prometheus metrics and what they are used for.

FAQ

What is the difference between a Cactus Node and a Cactus API Server?

The node is what has an identity within your PKI and can be made up of 1-N API server instances that all share the same configuration/identity of the node. See deployment scenarios above for a much more detailed explanation.

Is the API server horizontally scalable?

Yes, 100%. Keep in mind though that the API server can be loaded up with arbitrary plugins meaning that if you write a plugin that has a central database that can only do 1 transaction per second, then it will not help you much that the API server itself is horizontally scalable because deploying a thousand instances of the API server will just result in you having a thousand instances of your plugin all waiting for that underlying database with its 1 TPS throughput hogging your system. When we say that the API server is horizontally scalable, we mean that the API server itself is designed not to have any such state mentioned in the example above. You are responsible for only deploying plugins in the API server that are horizontally scalable as well. In short, your whole system is only horizontally scalable if all components of it are horizontally scalable.

Does the API server automatically protect me from malicious plugins?

No. If you install a third-party plugin that wasn’t vetted by anyone and that plugin happens to have malicious code in it to steal your private keys, it can do so. You are responsible for making sure that the plugins you install have no known security vulnerabilities or backdoors e.g. they are considered “secure”. The double quotes around “secure” is meant to signify the fact that no software is ever really truly secure, just merely lacking of known vulnerabilities at any given point in time.

Can I use the API server with plugins deployed as a service?

Yes. You can deploy your plugin written in any language, anywhere as long as it is accessible over the network and does come with a Typescript API client that you can use to install into the API server as a proxy for an in-process plugin implementation.

Previous Next