kathan24 / rtds-test

Envoy RTDS Prototype

This repo is forked from Ivan Han's rtds-test. I have extended it to prototype

• RTDS changes propagated to entire cluster rather than individual node.
• Clear runtime value from Envoy sidecar once the purpose is fulfilled.

Prerequisites

• You will need the Envoy binary.

Steps

• Clone this repo.

• Launch three Envoy proxies (in different terminals), pointing it to the config found in this repo. E.g. assuming your Envoy binary, envoy, is located under ~/Downloads, do:

  ~/Downloads/envoy -c $GOPATH/src/github.com/kathan24/rtds-test/config/envoy_google.yaml --service-node node1 --service-cluster cluster1
  
  ~/Downloads/envoy -c $GOPATH/src/github.com/kathan24/rtds-test/config/envoy_facebook.yaml --service-node node1 --service-cluster cluster2
  
  ~/Downloads/envoy -c $GOPATH/src/github.com/kathan24/rtds-test/config/envoy_microsoft.yaml --service-node node2 --service-cluster cluster2

• This will run two clusters - cluster1 and cluster2.

  • cluster1 has one node, node1
  • cluster2 has two nodes, node1 and node2

• Verify that Envoy sidecar works as expected

  • cluster1 with node1
    • http://localhost:10000 - redirects-to / loads https://www.google.com
    • http:localhost:9000 - you should see empty runtime values for Envoy
  • cluster2 with node1
    • http://localhost:10001 - redirects-to / loads https://www.facebook.com
    • http:localhost:9001 - you should see empty runtime values for Envoy
  • cluster2 with node2
    • http://localhost:10002 - redirects-to / loads https://www.microsoft.com
    • http:localhost:9002 - you should see empty runtime values for Envoy

• Launch RTDS management server

  go run $GOPATH/src/github.com/kathan24/rtds-test/main.go

• You should see

  • Within 10 seconds
    • cluster1 with node1 should have NO effect
    • cluster2 with node1 and node2
      • http://localhost:10001 and http://localhost:10002 - you should see fault filter abort with HTTP status 404
      • http://localhost:9001 and http://localhost:9002 - you should see runtime value as

        {
            "fault.http.abort.abort_percent": {
                "layer_values": [
                    "100"
                ],
                "final_value": "100"
            },
            "fault.http.abort.http_status": {
                "layer_values": [
                    "404"
                ],
                "final_value": "404"
            }
        }
        

  • After 10 seconds
    • Fault will be cleared and you should see the redirect happeneing again to www.facebook.com and www.microsoft.com for node1 and node2 of cluster2 respectively.