Understanding Load Balancer limits with graphql / websocket subscriptions
I have kubernetes cluster set up that hosts graphql api server https://hasura.io. This cluster is exposed to “the world” via digital ocean’s load balancer service.
After reading through load balancer limitations documentation I am concerned about two specific ones: 60s keep alive limit and max of 10,000 concurrent connections.
As my graphql api server exposes subscriptions through web-sockets and is used to power mobile game that relies on this live data, these limits can be reached relatively fast.
Hence I wanted to clarify if I understand these limits correctly and if they are indeed a concern for my use case. If so are there any alternatives to Load Balancers that I could utilise in this scenario?