Crystal WebSockets demo sample for Cloud Run

kemal-redis-chat

This sample demonstrates how to use WebSockets on Cloud Run with Crystal. The demo is a web app, that supports a single chat room. Crystal is a compiled language that has a syntax similar to Ruby. It is known to be very fast (as fast as Go), and the binaries are very small. The web server is known as Kemal, and supports WebSockets out of the box.

In order to have a consistent chat message queue, especially when there are multiple web servers due to auto-scaling, we must use a database for persistence. We use Redis, since it is fast and simple to set-up.

• Crystal Websockets sample for Cloud Run
  • Running locally
  • Deploying to Cloud Run
  • Deploying via Cloud Run button

Running locally

Before you can run the code locally, you need to do the following:

1. Install crystal and its dependency manager shards.

2. Install dependencies:

   With shards:

    shards install
   

3. Run locally

   With crystal:

    crystal run src/kemal-redis-chat.cr
   

[!NOTE] Unless you have a local redis DB running, the program will crash.

Deploying to Cloud Run

1. Create a Redis instance on Cloud Memorystore. Make sure to choose the VPC network you will use (default). After it’s created, note its IP address.

   export REGION=us-central1
   
   gcloud redis instances create redis-chat --size=1 --region=$REGION \
   --connect-mode=PRIVATE_SERVICE_ACCESS
   

2. Deploy to Cloud Run, with Direct VPC egress:

   export PROJECT_ID=$(gcloud config list --format 'value(core.project)' 2>/dev/null)
   export REDISHOST=$(gcloud redis instances describe redis-chat --region $REGION --format "value(host)")
   
   gcloud beta run deploy websockets \
       --source . \
       --allow-unauthenticated \
       --region $REGION \
       --max-instances 10 \
       --concurrency 100 \
       --timeout 3600 \
       --network=default \
       --subnet=default \
       --vpc-egress=private-ranges-only \
       --set-env-vars REDIS=${REDISHOST} \
       --set-env-vars DEBUG="false"
   

Deploying via Cloud Run button

Click on the button below to:

1. Launch a Cloud Shell machine
2. Download this GitHub repository
3. Build a container using Docker
4. Push the container into Google Artifact Registry
5. Launch Cloud Run

All automatically...

[!IMPORTANT] Since this code requires a redis DB, and the appropriate ENV variable set, it will not work directly via the Cloud Run button. However, it should be easy to re-deploy a new revision with the proper ENV variables and network settings.

Describe Cloud Run service

gcloud beta run services describe websockets \
    --region=$REGION

Clean up

gcloud run services delete websockets --region $REGION
gcloud redis instances delete redis-chat --region $REGION

---

Development

I added a Makefile to simply building and deploying the executable. I also created a Docker compose file for local testing.

Available make commands

$ make
Available targets:

  build/cloud                         Build docker container in Cloud
  build/docker                        Build docker container locally
  build/local                         Compiles source using shards command
  clean/cloud                         Delete Cloud Run service and container
  deploy                              Deploy container to Cloud Run
  docker/login                        Configures Docker to authenticate to GCR
  help                                This help screen
  logs                                Examine the logs from the Cloud container
  logs/stream                         Stream Cloud Run logs
  watch                               Locally run program with dynamic recompile

Resources

Cloud Run

1. https://github.com/GoogleCloudPlatform/cloud-run-button

GNU Make

1. GNU make
2. The Art of Makefiles
3. Makefile Best Practices
4. Learn Makefiles

Notes

Send authorization token, if unauthenticated is not allowed

curl --header "Authorization: Bearer $(gcloud auth print-identity-token)" [URL]

It is now possible to send requests via a local proxy, even if the service is authenticated.

gcloud beta run services proxy

Running Redis locally

docker run --rm -d -p 6379:6379 --name myredis redis:alpine

Proxy into GCP Redis via jumphost

https://cloud.google.com/memorystore/docs/redis/connect-redis-instance?hl=en#connecting_from_a_local_machine_with_port_forwarding

gcloud compute instances create NAME --machine-type=f1-micro --zone=ZONE
gcloud compute ssh COMPUTE_VM_NAME --zone=ZONE -- -N -L 6379:REDIS_INSTANCE_IP_ADDRESS:6379

Appendix

• https://cloud.google.com/run/docs/tutorials/websockets
• https://github.com/GoogleCloudPlatform/nodejs-docs-samples/tree/main/run/websockets
• https://serdardogruyol.com/building-a-realtime-chat-app-with-crystal-and-kemal
• https://www.tlbx.app/blog/dispatch-kemal-websocket-messages-with-redis-pub-sub
• https://kemalcr.com/
• https://crystal-lang.org/