Implemented graphql gateway update interval

[hansehe]

The service works well, but I have a problem when backend services are updated and the graphql gateway doesn't fetch the new changes. This solution should simply solve the problem.

[jakubknejzlik]

Hello @hansehe, thanks for Your contribution. We are deploying this gateway usually in k8s or lambda and using following approach:

• k8s: run kubectl rollout restart deployment/my-fancy-gateway
• lambda: updating the description of lambda 😄

...so this approach is really welcome. Also there's a plan to create schema repository where You can push the schema from each service. Right now if one of the service is down during the initialization, it is basically removed from schema.

[jakubknejzlik]

Jobs for deployment has been restricted to tag/master only so the checks should not fail.

[jakubknejzlik]

@hansehe I'm checking the latest commit, but we probably not understood each other :)

What I meant by updating the middleware instance is:

const { getENV } = require("./env");
const express = require("express");
const { getApolloServer } = require("./apollo");

(async () => {
  const app = express();
  const jsonBodyLimit = getENV("GRAPHQL_JSON_BODY_LIMIT", "2mb");
  app.use(express.json({ limit: jsonBodyLimit }));

  const server = await getApolloServer();
  let middleware = server.getMiddleware({});
  app.use((req, res, next) => {
    middleware(req, res, next);
  });

  const activateUpdateGatewayInterval = getENV("GRAPHQL_UPDATE_GATEWAY", "false") === "true";
  const updateGatewayInterval = getENV("GRAPHQL_UPDATE_GATEWAY_INTERVAL_MS", "60000");
  if (activateUpdateGatewayInterval === true)
  {
    setInterval(async () => {
      try {
        const server = await getApolloServer();
        middleware = server.getMiddleware({}); // just update the already existing instance
      } catch (error) {
        console.error(error);
      }
    }, updateGatewayInterval);
  }

  const HOST = getENV("HOST", "http://localhost");
  const PORT = getENV("PORT", "80");

  app.listen({ port: PORT }, () => {
    console.log(`🚀 Server ready at ${HOST}:${PORT}/graphql`);
  });
})();

module.exports.getServer = getApolloServer;

[hansehe]

Ok, thanks again! I updated the code with your reviews and it works with my local testing.

[jakubknejzlik]

Perfect, lookgs ok! Thanks for Your patience :)

[jakubknejzlik]

Hello @hansehe again, sorry for bringing this up, but I just noticed that there's experimental_pollInterval. Although there's this thread: https://github.com/apollographql/apollo-server/discussions/4280 I believe switching to it could be easier. What do You think?