Skip to content

Latest commit

 

History

History
150 lines (111 loc) · 5.8 KB

README.md

File metadata and controls

150 lines (111 loc) · 5.8 KB
Raw

Sample UI

This UI provides visualization of resources and interface to deploy connectivity workflows.

It allows you to:

  1. discover network domains
  2. create new network domains (stub only)
  3. create connections between two network domains
  4. view connections between network domains
  5. create SLA profiles for aforementioned connections

Tech stack

  1. React + TypeScript, redux-toolkit
  2. CRA + Craco
  3. Ant Design, ag-grid community
  4. React Router
  5. grpc

How to run

You can decide whether the app should run on stub data or on live vmanage servers. You set this in your local .env.local file in the root. If REACT_APP_BUILD is set to live the app will try to fetch all the data from vmanage and if it's set to stub it will try to access json-server with data in src/_db/db.json. In both cases you should follow the steps bellow.

  1. Run npm install from the root folder
  2. Setup vmanage servers and Envoy proxy
  3. Setup REACT_APP_BUILD to desired setting
  4. To start the UI run npm start from the root. This starts both the actual client and stubbed data from json-server

UI should be available at http://localhost:5001/

Building docker image

To build your image simply run

docker build -t sample-ui:1.0 .

To push your image to ECR registry tag it and then push

docker tag sample-ui:1.0 229451923406.dkr.ecr.us-west-2.amazonaws.com/awi/sample-ui:1.0
docker push 229451923406.dkr.ecr.us-west-2.amazonaws.com/awi/sample-ui:1.0

The current image uses nginx as a base which helped decreasing the image size to 90 MB and also made it much faster to start the container as there is no building process during the start but rather serving static files.

Backend Calls

It is crucial to remember that JS calls implemented in this UI are not sent by the UI process/container but they are coming directly from the web browser or any other client used by the user.

JS Calls Image

In this example, the UI application serves a JS file with a logic that performs a call to the backend service present within the same cluster, reachable from the UI on port :8000. The problem is that the call will be performed by the Chromium browser used by the client and so it will try to send it to the localhost on port 8000 which will most likely fail (unless the user uses the browser on the same host where the backend server runs).

So the JS files cannot use the addressing used internally within a server machine or a k8s cluster - these addresses must be reachable from the user's device.

Of course, the backend service does not need to be exposed directly but it will be rather hidden behind a reverse proxy. It, however, gives us another problem - the front-end application needs to be aware of the public address which is used by the proxy. It can be solved by a few options:

  • using final DNS name - having an actual DNS name associated with our server, we can simply hardcode it. However, it is not a good solution as changing the domain will require rebuilding the front-end application. Additionally, this solution works only if it is expected that there will be only one instance of front-end (SaaS) but if each potential customer can have his own front-end instance deployed on on-prem systems, it becomes a problem (even in case of SaaS we still may want to deploy separate production and testing front-end)

  • providing address as a runtime parameter - the second option is to use a variable with the address which is set with the environment variable or commandline argument. That way the entire system can be deployed, the public address retrieved and then the UI application restarted with modified config (config-map in k8s deployment or docker-compose scripts). However, using variables requires JS server which results in higher image size and container start-up time. Serving static JS files with NGinx image saves a lot of size and gives a quick application restart - however it makes it harder to use a configuration variable - such solution would require either some nginx libraries/extensions or simple tools to override every JS file using that address with a proper value (simple sed script for instance). This solution is also problematic if our public address changes quite often - it would require constant front-end restarting.

  • do not provide an actual address

The last solution is the one we use:

JS Proxying Image

The approach is simple. The JS code describes only endpoints for backend calls, for instance /grpc/Clouds rather than http://myapp.net.io/grpc/Clouds. If the address is not provided, the client will use the address it used to download a particular JS Document.

For instance, we have entered our UI by specifying 127.0.0.1:8080/index.php in the web browser. The page defines one script that calls backend service using /grpc call. Since the index.php page was downloaded from 127.0.0.1:8080 the same address will be used to complete /grpc call giving as a result the address 127.0.0.1:8080/grpc.

This solution has a condition - both UI and the backend need to be reachable under the same address. For that matter, we need a reverse proxy to push our front-end requests to the UI and our backend requests to corresponding services. The front-end cannot be in front as the logic above would redirect all backend calls back to the UI.

For that reason, we have introduced a /grpc prefix for all GRPC backend calls so that we can configure Envoy Proxy to distinguish backend requests starting with /grpc prefix from the rest of the calls that should be redirected to the UI.

Contributing

Thank you for interest in contributing! Please refer to our contributing guide.

License

sample-ui is released under the Apache 2.0 license. See LICENSE.