Skip to content

Latest commit

 

History

History
227 lines (167 loc) · 7.71 KB

README.md

File metadata and controls

227 lines (167 loc) · 7.71 KB

🌀hono-under-pressure🌀

tests npm version npm downloads license

Measure process load with automatic handling of "Service Unavailable" plugin for Hono. It can check maxEventLoopDelay, maxHeapUsedBytes, maxRssBytes and maxEventLoopUtilization values. You can also specify a custom health check, to verify the status of external resources.

Install

# Using npm/yarn/pnpm/bun
npm add hono-under-pressure

Usage

Wrap the function around your server instance and the provided middlewares can be added to Hono instance to apply them.

import { underPressure } from "hono-under-pressure";
import { serve } from "@hono/node-server";
import { Hono } from "hono";

const app = new Hono();
app.get("/", (c) => {
  const isUnderPressure = c.get("isUnderPressure");
  if (isUnderPressure()) {
    // skip complex computation
  }
  return c.text("Hello Node.js!");
});

underPressure(
  (handlers) => {
    const newApp = new Hono().use(...handlers);
    newApp.route("/", app);
    return serve(newApp);
  },
  {
    maxEventLoopDelay: 1000,
    maxHeapUsedBytes: 100000000,
    maxRssBytes: 100000000,
    maxEventLoopUtilization: 0.98,
  }
);

hono-under-pressure will automatically handle for you the Service Unavailable error once one of the thresholds has been reached. You can configure the error handler by passing pressureHandler function.

underPressure(createServer, {
  pressureHandler: () => {
    throw new HTTPException(SERVICE_UNAVAILABLE, {
      message: "Under pressure!",
    });
  },
});

The default value for maxEventLoopDelay, maxHeapUsedBytes, maxRssBytes and maxEventLoopUtilization is 0. If the value is 0 the check will not be performed.

Since eventLoopUtilization is only available in Node version 14.0.0 and 12.19.0 the check will be disabled in other versions.

memoryUsage

This plugin also exposes a function that will tell you the current values of heapUsed, rssBytes, eventLoopDelay and eventLoopUtilized.

const memoryUsage = c.get("memoryUsage");
console.log(memoryUsage());

Pressure Handler

You can provide a pressure handler in the options to handle the pressure errors. The advantage is that you know why the error occurred. Moreover, the request can be handled as if nothing happened.

underPressure(createServer, {
  maxHeapUsedBytes: 100000000,
  maxRssBytes: 100000000,
  pressureHandler: (c, type, value) => {
    if (type === underPressure.TYPE_HEAP_USED_BYTES) {
      console.warn(`too many heap bytes used: ${value}`);
    } else if (type === underPressure.TYPE_RSS_BYTES) {
      console.warn(`too many rss bytes used: ${value}`);
    }

    throw new HTTPException(503, { message: "out of memory" }); // if you omit this line, the request will be handled normally
  },
});

It is possible as well to return a Promise that will call c.text (or something else).

underPressure(createServer, {
  maxHeapUsedBytes: 100000000,
  pressureHandler: (c, type, value) => {
    return getPromise().then(() => {
      throw new HTTPException(503, { message: "out of memory" });
    });
  },
});

If you don't throw a HTTPException, the request will be handled normally.

It's also possible to specify the pressureHandler on the route:

import { underPressure } from "hono-under-pressure";
import { serve } from "@hono/node-server";
import { Hono } from "hono";

const app = new Hono();

app.use((c, next) => {
  c.set("pressureHandler", (c, type, value) => {
    if (type === underPressure.TYPE_HEAP_USED_BYTES) {
      console.warn(`too many heap bytes used: ${value}`);
    } else if (type === underPressure.TYPE_RSS_BYTES) {
      console.warn(`too many rss bytes used: ${value}`);
    }

    throw new HTTPException(503, { message: "out of memory" }); // if you omit this line, the request will be handled normally
  });
});

app.get("/", (c) => {
  return c.text("A");
});

underPressure(
  (handlers) => {
    const newApp = new Hono().use(...handlers);
    newApp.route("/", app);
    return serve(newApp);
  },
  {
    maxHeapUsedBytes: 100000000,
    maxRssBytes: 100000000,
  }
);

Custom health checks

If needed you can pass a custom healthCheck property, which is an async function, and hono-under-pressure will allow you to check the status of other components of your service.

This function should return a promise that resolves to a boolean value or to an object. The healthCheck function can be called every X milliseconds, the time can be configured with the healthCheckInterval option.

By default when this function is supplied your service health is considered unhealthy, until it has started to return true.

underPressure(createServer, {
  healthCheck: async function () {
    // do some magic to check if your db connection is healthy, etc...
    return true;
  },
  healthCheckInterval: 500,
});

Sample interval

You can set a custom value for sampling the metrics returned by memoryUsage using the sampleInterval option, which accepts a number that represents the interval in milliseconds.

The default value is different depending on which Node version is used. In version 8 and 10 it is 5, while on version 11.10.0 and up it is 1000. This difference is because from version 11.10.0 the event loop delay can be sampled with monitorEventLoopDelay and this allows to increase the interval value.

underPressure(
  createServer,
  {
    sampleInterval: <your custom sample interval in ms>
  }
);

Additional information

setTimeout vs setInterval

Under the hood the hono-under-pressure uses the setTimeout method to perform its polling checks. The choice is based on the fact that we do not want to add additional pressure to the system.

In fact, it is known that setInterval will call repeatedly at the scheduled time regardless of whether the previous call ended or not, and if the server is already under load, this will likely increase the problem, because those setInterval calls will start piling up. setTimeout, on the other hand, is called only once and does not cause the mentioned problem.

One note to consider is that because the two methods are not identical, the timer function is not guaranteed to run at exactly the same rate when the system is under pressure or running a long-running process.

Articles

These can help you understand how it can improve your Hono Nodejs server performance -

Contributing

We would love to have more contributors involved!

To get started, please read our Contributing Guide.

Credits

The hono-under-pressure project is heavily inspired by @fastify/under-pressure