Use next server when maxconn is reached

I’ve found that with most load balancing algorithms, if the selected server has reached maxconn then additional requests will be queued on that server. balance first is an exception – if the first server has reached maxconn then it moves on to the next one. Is it possible to have other algorithms do the same thing? I’d like to use consistent hashing with long running (WebSocket) connections. Currently, if a request gets placed on a server that has reached maxconn then it gets queued, but since these requests are long-lived it’s unlikely to reach the server before timeout queue is reached.

backend back
  balance uri
  timeout queue 5s
  server srv1 127.0.0.1:8001 maxconn 10
  server srv2 127.0.0.1:8002 maxconn 10

1 post - 1 participant

Read full topic