groupcache-example

An example service using groupcache as a distributed in memory cache.

Running

minikube

1. run make run (note: starts a continuous process in the terminal for a tunnel to the deployment)

🏃  Starting tunnel for service groupcache-svc.
|--------------------|----------------|-------------|------------------------|
|     NAMESPACE      |      NAME      | TARGET PORT |          URL           |
|--------------------|----------------|-------------|------------------------|
| groupcache-example | groupcache-svc |             | http://127.0.0.1:50899 |
|--------------------|----------------|-------------|------------------------|
http://127.0.0.1:50899
❗  Because you are using a Docker driver on darwin, the terminal needs to be open to run it.

2. grab the URL from the run command, e.g. http://127.0.0.1:50899
3. send a request to the service:

$ curl "http://127.0.0.1:50899/data/some-id`
{"guid":"some-id","date_created":"2021-11-17T19:39:37.0937256Z"}

4. View application logs: make minikube-logs
5. run make clean to tear down the services

docker-compose

A docker-compose.yml is provided to run the services locally. This will start three instances of a server that can be used to fetch data from a backend and an nginx proxy to front them. The nginx proxy provides a single host to call to load balance across the running instances. This is done to see the effects of groupcache in action.

1. run docker-compose up
2. send a request to the service (note: the host will always be on port 8080)

$ curl "http://127.0.0.1:8080/data/some-id"
{"guid":"hi","date_created":"2021-11-17T19:44:52.7587009Z"}

3. run docker-compose down to tear down the services

Sending Requests

Send a request to the nginx (if using docker-compose) service which will be proxied to one of the api services.

$ curl "http://localhost:8080/data/U4JcsWos6U3sgfYaq1vhI8ssln0wTmys"
{
  "guid": "fupPlyUOYXUgeSupjwQNhnR4W666oqkv",
  "date_created": "2021-11-02T23:48:52.1826762Z"
}

server-03_1  | 2021/11/02 23:48:52 fetching key from backend: fupPlyUOYXUgeSupjwQNhnR4W666oqkv
server-03_1  | 23:48:51 | 200 |      0s |      172.31.0.3 | GET     | /_groupcache/data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv
server-01_1  | 23:48:51 | 200 |     3ms |      172.31.0.5 | GET     | /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv
nginx_1      | 172.31.0.1 - - [02/Nov/2021:23:48:52 +0000] "GET /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv HTTP/1.1" 200 89 "-" "Paw/3.3.1 (Macintosh; OS X/11.6.0) GCDHTTPRequest" "-"
server-03_1  | 23:48:56 | 200 |      0s |      172.31.0.2 | GET     | /_groupcache/data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv
server-02_1  | 23:48:56 | 200 |     2ms |      172.31.0.5 | GET     | /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv
nginx_1      | 172.31.0.1 - - [02/Nov/2021:23:48:56 +0000] "GET /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv HTTP/1.1" 200 89 "-" "Paw/3.3.1 (Macintosh; OS X/11.6.0) GCDHTTPRequest" "-"
server-03_1  | 23:48:56 | 200 |     1ms |      172.31.0.5 | GET     | /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv
nginx_1      | 172.31.0.1 - - [02/Nov/2021:23:48:56 +0000] "GET /data/fupPlyUOYXUgeSupjwQNhnR4W666oqkv HTTP/1.1" 200 89 "-" "Paw/3.3.1 (Macintosh; OS X/11.6.0) GCDHTTPRequest" "-"

Above are three different requests all for the same key.

• the first goes to server-01_1 which makes a request to server-03 (the owner of the key). server-03 does not have the key in the cache so it looks it up in the backend
• the second goes to server-02 which makes a request to server-03 to get the data
• the third goes to server-03 which is the owner of the key, but the item is cached from the first call, so it returns immediately.

wrk2

a script (requests.wrk.lua) is provided to use with wrk2 to generate load to see the service in action.

To install wrk2 do the following:

$ brew tap jabley/homebrew-wrk2
$ brew install --HEAD wrk2

This script randomly generates ids to fetch from 1-100 to simulate a variety of hot links. To run the script, specify it as the script arg to wrk2. For example:

$ wrk2 --rate 250 --connections 10 --duration 15 --threads 2 --script requests.wrk.lua http://127.0.0.1:62836

note: replace the url with the url for the service (either from the minikube tunnel or the docker-compose service)