Redis instance is being mapped only once per app visit and only to the page that is being hit

Prerequisites

• I have written a descriptive issue title
• I have searched existing issues to ensure the issue has not already been raised

Issue

Reproducible demo/repo here

I am mapping my redis instance to req.raw, but for some reason, it doesn’t persist through page navigations, even if you return to the same page.

In my afterHandler, I’m doing the following:

fastify.next('/', (app, req, reply) => {
	req.raw.redisInstance = fastify.redis;
	app.render(req.raw, reply.raw, '/', req.query);
});

fastify.next('/another', (app, req, reply) => {
	req.raw.redisInstance = fastify.redis;
	app.render(req.raw, reply.raw, '/another', req.query);
});

If you visit the index file or the another route, then the redis instance will be available the first time. The redis instance is also available if you refresh the page.

However if you first first index, then go to another, and then go back to index, the reference to the redis instance is lost.

I’ve tried doing as was suggested here, but to no avail.

What am I doing wrong here?

[radomird]

After looking into the suggestion that @climba03003 proposed, I’ve realised that there could be a very easy solution to this issue which doesn’t require a change in the plugin.

@mkhoussid: In order to achieve to have the redis instance on all of the requests you could simply just add a onRequest hook and move the logic from your afterHandler to the hook and that will solve the problem.

fastify.addHook("onRequest", (req, reply, done) => {
  req.raw.redisInstance = fastify.redis;
  done();
});

Here is the full code (from your example):

server.js:

const fastifyConstructor = require("fastify");
const nextJs = require("fastify-nextjs");

const PORT = process.env.PORT || 3000;
const isDev = process.env.NODE_ENV !== "production";

const { registerPlugins, afterHandler, listenHandler } = require("./utils");

console.log("Environment: ", process.env.NODE_ENV);
const fastify = fastifyConstructor({
  // logger: isDev,
});

registerPlugins(fastify, isDev);

fastify.register(nextJs, { dev: isDev }).after(afterHandler(fastify, isDev));

fastify.addHook("onRequest", (req, reply, done) => {
  req.raw.redisInstance = fastify.redis;
  done();
});

fastify.listen(PORT, "::", listenHandler(PORT));

afterHandler.js

module.exports = (fastify) =>
  async function () {
    try {
      fastify.next('/');
      fastify.next('/another');
    } catch (err) {
      console.log('Error caught in afterHandler', err);
    }
};

[mkhoussid]

After looking into the suggestion that @climba03003 proposed, I’ve realised that there could be a very easy solution to this issue which doesn’t require a change in the plugin.

@mkhoussid: In order to achieve to have the redis instance on all of the requests you could simply just add a onRequest hook and move the logic from your afterHandler to the hook and that will solve the problem.

fastify.addHook("onRequest", (req, reply, done) => {
  req.raw.redisInstance = fastify.redis;
  done();
});

Here is the full code (from your example):

server.js:

const fastifyConstructor = require("fastify");
const nextJs = require("fastify-nextjs");

const PORT = process.env.PORT || 3000;
const isDev = process.env.NODE_ENV !== "production";

const { registerPlugins, afterHandler, listenHandler } = require("./utils");

console.log("Environment: ", process.env.NODE_ENV);
const fastify = fastifyConstructor({
  // logger: isDev,
});

registerPlugins(fastify, isDev);

fastify.register(nextJs, { dev: isDev }).after(afterHandler(fastify, isDev));

fastify.addHook("onRequest", (req, reply, done) => {
  req.raw.redisInstance = fastify.redis;
  done();
});

fastify.listen(PORT, "::", listenHandler(PORT));

afterHandler.js

module.exports = (fastify) =>
  async function () {
    try {
      fastify.next('/');
      fastify.next('/another');
    } catch (err) {
      console.log('Error caught in afterHandler', err);
    }
};

Yes, using hooks is also a valid approach. It needs to be tested and documented.

@radomird Could you send a PR for the test-case and documentation?

Works.

Sorry @radomird , docs are great, but writing them sucks.