Load reporting service (LRS)

This simple example demonstrates Envoy’s Load reporting service (LRS) capability and how to use it.

Lets say Cluster A (downstream) talks to Cluster B (Upstream) and Cluster C (Upstream).

When enabling Load Report for Cluster A, the LRS server should be sending LoadStatsResponse to Cluster A with LoadStatsResponse.Clusters to be B and C.

The LRS server will then receive LoadStatsRequests (with total requests, successful requests etc) from Cluster A to Cluster B and from Cluster A to Cluster C.

In this example, all incoming requests are routed via Envoy to a simple Golang web server named http_server.

We scale up two containers and randomly send requests to each.

Envoy is configured to initiate the connection with the LRS Server.

The LRS Server enables stats by sending LoadStatsResponse.

Sending requests to http_server are counted towards successful requests and will be visible in the LRS Server logs.

Step 1: Build the sandbox

Change to the load-reporting-service directory.

Terminal 1

$ pwd
examples/load-reporting-service
$ docker compose pull
$ docker compose up --scale http_service=2

Terminal 2

$ pwd
examples/load_reporting_service
$ docker compose ps

              Name                               Command               State                Ports
-----------------------------------------------------------------------------------------------------------------
load-reporting-service_http_service_1   /docker-entrypoint.sh /usr ... Up           10000/tcp, 0.0.0.0:81->80/tcp
load-reporting-service_http_service_2   python3 /code/service.py   ... Up (healthy)
load-reporting-service_lrs_server_1     go run main.go                 Up

Step 2: Start sending stream of HTTP requests

Terminal 2

$ pwd
examples/load_reporting_service
$ bash send_requests.sh

The script above (send_requests.sh) sends requests randomly to each Envoy, which in turn forwards the requests to the backend service.

Step 3: See Envoy Stats

You should see

Terminal 1

............................
lrs_server_1    | 2020/02/12 17:08:20 LRS Server is up and running on :18000
lrs_server_1    | 2020/02/12 17:08:23 Adding new cluster to cache `http_service` with node `0022a319e1e2`
lrs_server_1    | 2020/02/12 17:08:24 Adding new node `2417806c9d9a` to existing cluster `http_service`
lrs_server_1    | 2020/02/12 17:08:25 Creating LRS response for cluster http_service, node 2417806c9d9a with frequency 2 secs
lrs_server_1    | 2020/02/12 17:08:25 Creating LRS response for cluster http_service, node 0022a319e1e2 with frequency 2 secs
http_service_2  | 127.0.0.1 - - [12/Feb/2020 17:09:06] "GET /service HTTP/1.1" 200 -
http_service_1  | 127.0.0.1 - - [12/Feb/2020 17:09:06] "GET /service HTTP/1.1" 200 -
............................
lrs_server_1    | 2020/02/12 17:09:07 Got stats from cluster `http_service` node `0022a319e1e2` - cluster_name:"local_service" upstream_locality_stats:<locality:<> total_successful_requests:21 total_issued_requests:21 > load_report_interval:<seconds:1 nanos:998411000 >
lrs_server_1    | 2020/02/12 17:09:07 Got stats from cluster `http_service` node `2417806c9d9a` - cluster_name:"local_service" upstream_locality_stats:<locality:<> total_successful_requests:17 total_issued_requests:17 > load_report_interval:<seconds:1 nanos:994529000 >
http_service_2  | 127.0.0.1 - - [12/Feb/2020 17:09:07] "GET /service HTTP/1.1" 200 -
http_service_1  | 127.0.0.1 - - [12/Feb/2020 17:09:07] "GET /service HTTP/1.1" 200 -
............................
lrs_server_1    | 2020/02/12 17:09:09 Got stats from cluster `http_service` node `0022a319e1e2` - cluster_name:"local_service" upstream_locality_stats:<locality:<> total_successful_requests:3 total_issued_requests:3 > load_report_interval:<seconds:2 nanos:2458000 >
lrs_server_1    | 2020/02/12 17:09:09 Got stats from cluster `http_service` node `2417806c9d9a` - cluster_name:"local_service" upstream_locality_stats:<locality:<> total_successful_requests:9 total_issued_requests:9 > load_report_interval:<seconds:2 nanos:6487000 >