Best way to handle REST requests with external programs?-CodePudding

TL;DR
How can I setup a light-weight web server to execute external programs to handle REST requests?

The long version:
We have a set of services and databases deployed in Kubernetes via Helm. There are some executables that perform maintenance, cleanup, backup, restore etc that I need to execute (some on-demand & some periodically).
I want to park a small, light-weight web server somewhere mounted with access to the binaries and execute them when REST requests are handled.

server needs to have a small memory footprint
traffic will be really light (like minutes between each request)
security is not super important (it will run inside our trusted zone)
server needs to handle GET and POST (i.e. passing binary content TO & FROM external program)

I've glanced at lighttpd or nginx with CGI modules but I'm not experienced with those.
What do you recommend? Do you have a small example to show how to do it?

CodePudding user response：

Here's k8s native approach:

... a set of services and databases deployed in Kubernetes... some executables that perform maintenance, cleanup, backup, restore etc...some on-demand & some periodically

If you can bake those "executables" into an image, you can run these programs on-demand as k8s job, and schedule repeating job as k8s cronjob. If this is possible in your context then you can create a k8s role that has just enough right to call job/cronjob api, and bind this role to a dedicated k8s service account.

Then you build a mini web application using any language/framework of your choice, run this web application on k8s using the dedicated service account, expose your pod as service using NodePort/LoadBalancer to receive GET/POST requests. Finally you make direct api call to k8s api-server to run jobs according to your logic.

CodePudding user response：

I think I'd make a small service. I'd choose Node, since that's something I'm familiar with, but lots of languages have adopted this way of setting up a server, so choose something you like.

index.js

const express = require('express')
const { exec } = require('child_process')
const app = express()
const port = 3000

app.get('/os-release', (req, res) => {
  exec('cat /etc/os-release', (error, stdout, stderr) => {
    res.send(stdout)
  })
})

app.listen(port, () => {
  console.log(`Listening on port ${port}`)
})

package.json

{
  "name": "shell-exec",
  "version": "1.0.0",
  "description": "",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "author": "",
  "license": "ISC",
  "dependencies": {
    "child_process": "^1.0.2",
    "express": "^4.17.1"
  }
}

Dockerfile

FROM node:alpine
EXPOSE 3000
WORKDIR /app
COPY package.json ./
RUN npm install
COPY . ./
CMD ["node", "index.js"]

I should be relatively easy to make POST requests that take input parameters if you need to do that. I also like that you can send the output of the command back to the caller.

If you build and run this, you can then hit the /os-release endpoint and get data back from the cat /etc/os-release command.