Skip to main content

server

Core tramvai module, responsible for processing the users requests.

Installation

You need to install @tramvai/module-server

npm i --save @tramvai/module-server

And connect to the project

import { createApp } from '@tramvai/core';
import { ServerModule } from '@tramvai/module-server';

createApp({
name: 'tincoin',
modules: [ServerModule],
});

Explanation

Processing the users requests

ServerModule creates express.js application, handles user requests, runs commandLineRunner, and sends responses to users with data, headers and status from RESPONSE_MANAGER_TOKEN token.

Request proxying

ServerModule allows you to configure the proxying of urls to the application using the library http-proxy-middleware

To enable proxying, create a file proxy.conf.js or proxy.conf.json in the root of the project to export the request mapping object, or you can use the PROXY_CONFIG_TOKEN token.

Proxy config format

Key-value object
const testStand = 'https://example.org';

module.exports = {
// The key is the path pattern for the `express` to be passed to `app.use`
// value can be a string, in order to proxy all urls starting with `/login/`
'/login/': testStand,
// or can be a config object for [http-proxy](https://github.com/chimurai/http-proxy-middleware#http-proxy-options)
'/test/': {
target: testStand,
auth: true,
xfwd: true,
...
}
};
Object with context and target properties
module.exports = {
// context - is similar to the option for [http-proxy-middleware](https://github.com/chimurai/http-proxy-middleware#context-matching)
context: ['/login/', '/registration/', '/auth/papi/'],
target: 'https://example.org',
// other `http-proxy-middleware` options
changeOrigin: true,
};
Array with context and target properties
[
{
"context": ["/a/", "/b/*/c/"],
"target": "https://example.org"
}
]
Implementation of the PROXY_CONFIG_TOKEN token
import { Scope, provide } from '@tramvai/core';
import { PROXY_CONFIG_TOKEN } from '@tramvai/tokens-server';

[
provide({
provide: PROXY_CONFIG_TOKEN,
scope: Scope.SINGLETON,
useValue: {
context: ['/a/', '/b/*/c/'],
target: 'https://example.org',
},
multi: true,
}),
];

Serving static files

The ServerModule has a built-in static server that allows you to distribute static files to users.

To serve files, you need to create a directory public in the root of the project in which to place the necessary files. After that, all files will be available for request by browsers.

For example, we want to distribute sw.js file from the project's root: for this we create a folder public in which we put the file sw.js. Now on the client side, we will be able to request data from the url http://localhost:3000/sw.js. Also, we will most likely need some modifications on the CI/CD side to copy the public folder to the stands.

This function is also available in production. For this purpose, copy the folder public into the docker container

caution

Module will not serve the newly added file on the filesystem, works only for all defined files in the served folder at the time of the server startup.

PAPI

Papi - API routes for the tramvai application. More information is available in Papi

Emulation of network/backends problems in the application

(functionality is only available in dev mode)

The server has the ability to increase the response time of all requests.

To do this you must:

  • start the application
  • send a POST request to /private/papi/debug-http-request with a delay for the request:
curl --location --request POST 'http://localhost:3000/tincoin/private/papi/debug-http-request' \
--header 'Content-Type: application/x-www-form-urlencoded' \
--data-urlencode 'delay=2000'
  • check if the application works. Note: after each restart of the server the settings are reset, so after each rebuild it is necessary to access papi again.
  • you can disable the timeout by accessing the same papi using the DELETE method
curl --location --request DELETE 'http://localhost:3000/tincoin/private/papi/debug-http-request'

Logging requests sent to the server

In dev mode, all requests sent through the standard http and https libraries for nodejs are logged under a special server:node-debug:request key. This allows you to see all requests that have been sent to the server, even if no logging has been defined for the requests explicitly.

To enable such logging, simply add the server:node-debug:request key to the LOG_ENABLE environment variable

Health checks

  • /healthz - always replies OK after starting the application
  • /readyz - always replies OK after starting the application

Metrics

The metrics module is automatically connected into the server module. For more information on metrics, you can read in the metrics documentation

Warming application caches

The cache-warmup module is automatically plugged into the server module. Detailed information on cache warmup can be found in cache-warmup documentation

Custom headers

Building and Deployment Information

There are special headers in the module, which help to determine the exact information about the version of the built application, commit, branch, etc:

  • x-app-id - The name of the application specified in createApp. Specified in the application code.
  • x-host - Hostname of the server where the current application is running. Computed in runtime.
  • x-app-version - version of the running application. Transmitted through the environment variable APP_VERSION.
  • x-deploy-branch - branch from which the current application image was built. Passed through environment variable DEPLOY_BRANCH.
  • x-deploy-commit - sha commit from which current application image was built. Passed through environment variable DEPLOY_COMMIT.
  • x-deploy-version - deploy revision number in k8s. Passed through environment variable DEPLOY_VERSION.
  • x-deploy-repository - application repository link. Passed through environment variable DEPLOY_REPOSITORY.

For all of the headers above which are passed via environment variables to be available, you need the external infrastructure to pass them when building and deprovisioning the application image (inside tinkoff this is done automatically).

Server-Timing

Header Server-Timing provides additional information about different timings that were spent on server to handle current request. Currently, it shows timings for CommandLineRunner's lines execution.

To see values that related to request look for the header Server-Timing or check the Timing tab in browser DevTools for the page's main html.

Debugging

Module uses loggers with identifiers: server, server:static, server:webapp, server:node-debug:request

Early Hints

Module can send the 103 Early Hints response to provide better performance, though there are several limitations (look here and here).

If you want to enable Early Hints, provide EARLY_HINTS_ENABLED env variable:

EARLY_HINTS_ENABLED: 'true'

Or provide custom EARLY_HINTS_ENABLED_TOKEN token (EARLY_HINTS_ENABLED will be ignored):

import { EARLY_HINTS_ENABLED_TOKEN } from '@tramvai/tokens-server';

const provider = {
provide: EARLY_HINTS_ENABLED_TOKEN,
useValue: () => true,
};
info

You must check that the balancers and proxies in front of your application support Early Hints

Currently, module provides hints for next resources:

  • Resources with preconnectLink type;
  • General resources, which have the preloadLink type;
  • Page resources, that exist in the ResourcesRegistry with 'data-critical': "true" attribute;
  • Preconnects for CDN urls, that are computed from preloadLink or data-critical resources, described above;

Server will try to response with 103 as soon as possible and there can be more than one such responses.

curl -I -X HEAD http://localhost:3000

HTTP/1.1 103 Early Hints
Link: <https://www.cdn-tinkoff.ru>; rel=preconnect
Link: <https://www.cdn-tinkoff.ru/frontend-libraries/npm/react-kit-font/1.0.0/TinkoffSans.woff2>; rel=preload; as=font
Link: <https://www-stage.cdn-tinkoff.ru/frontend-libraries/feedback/1.14.0/feedback.css>; rel=preload; as=style
Link: <https://www-stage.cdn-tinkoff.ru>; rel=preconnect

HTTP/1.1 200 OK
...

How to

Setting keepAliveTimeout for the server

The default value for server's keepAliveTimeout is 5000. However, in case you want to set it manually just pass the environment variable NODE_KEEPALIVE_TIMEOUT=[your_value]. For further reading go to NodeJs server.keepAliveTimeout page.

Specify server port

By default server starts at 3000 port. You have next options to override this value depending on your environment:

  • in dev environment port in fully controlled by @tramvai/cli and should be specified by -p option
  • in prod environment it can be specified explicitly as an environment variable PORT e.g. PORT=8080

Specify port for utility paths

It includes: health checks, liveness checks, metrics.

By default port for utility paths equals base server port and they will use the same http server.

If you want to change this and run utility routes on different server with different port, use token UTILITY_SERVER_PORT_TOKEN:

import { UTILITY_SERVER_PORT_TOKEN } from '@tramvai/tokens-server';

const providers = [
{
provide: UTILITY_SERVER_PORT_TOKEN,
useValue: 6532,
},
];

Or use env variable UTILITY_SERVER_PORT with defined value.

env UTILITY_SERVER_PORT=6532 tramvai start app

Specify path for liveness and readiness probes

By default, liveness and readiness probes are available by healtz and readyz paths.

If you want to change this paths, use LIVENESS_PATH_TOKEN and READINESS_PATH_TOKEN tokens.

import { LIVENESS_PATH_TOKEN, READINESS_PATH_TOKEN } from '@tramvai/tokens-server';

const providers = [
{
provide: LIVENESS_PATH_TOKEN,
useValue: '/custom-liveness',
},
{
provide: READINESS_PATH_TOKEN,
useValue: '/custom-readiness',
},
];

Exportable tokens

Link