only one backend when loadbalancer.swarm=true

[siyu6974]

Do you want to request a feature or report a bug?

Bug

What did you do?

Running 2+ backends (Docker swarm 18.09) with the label traefik.backend.loadbalancer.swarm=true

What did you expect to see?

Get 1 frontend with 2+ backends

What did you see instead?

1 frontend with only 1 backend.

Output of traefik version: (What version of Traefik are you using?)

Version:      v1.7.4
Codename:     maroilles
Go version:   go1.11.1
Built:        2018-10-30_10

What is your environment & configuration (arguments, toml, provider, platform, ...)?

debug = false

logLevel = "ERROR"
defaultEntryPoints = ["https","http"]

[entryPoints]
  [entryPoints.http]
  address = ":80"
    [entryPoints.http.redirect]
    entryPoint = "https"
  [entryPoints.https]
  address = ":443"
  [entryPoints.https.tls]

[retry]

[docker]
endpoint = "tcp://127.0.0.1:2376"
domain = "my.domain"
watch = true
exposedByDefault = false
swarmMode = true

[acme]
email = "email@test.com"
storage = "acme.json"
entryPoint = "https"
onHostRule = true
[acme.httpChallenge]
entryPoint = "http"

If applicable, please paste the log output in DEBUG level (--logLevel=DEBUG switch)

This is the exact same issue as #2138, and yes I shamelessly copied "What did you expect to see" and "What did you see instead", however using the latest version, the issue is still there.

Everything is working when traefik.backend.loadbalancer.swarm is set to false.
When loadbalancer.swarm=true, the LB is actually working, when I make requests to the endpoint, by inspecting the log, I can see backends taking their turn logging the response.

[ldez]

Could you provide your docker-compose file?

Could you check that the labels are in the section deploy:
https://docs.traefik.io/master/configuration/backends/docker/#using-docker-with-swarm-mode

[siyu6974]

@ldez yes there are in deploy otherwise I believe it won't even work when docker.endpoint is set to port 2376.

version: '3'
services:
  backend:
    image: ${REGISTRY_URL}backend-${SUBDOMAIN}:${GIT_HASH:-latest}
    networks:
      traefik-net:
        aliases:
        - backend-${SUBDOMAIN}
    env_file: env_${SUBDOMAIN}/backend.env
    command: npm start
    deploy:
      labels:
      - "traefik.docker.network=traefik-net"
      - "traefik.frontend.rule=Host:api.${SUBDOMAIN}.${DOMAIN}"
      - "traefik.backend=loopback-${SUBDOMAIN}"
      - "traefik.backend.loadbalancer.swarm=true"
      - "traefik.backend.healthcheck.path=/"
      - "traefik.backend.healthcheck.interval=15s"
      - "traefik.backend.healthcheck.timeout=15s"
      - "traefik.port=3005"
      - "traefik.protocol=http"
      - "traefik.enable=true"
      update_config:
        parallelism: 2
        failure_action: rollback
        monitor: 10s
    ports:
    - "3005"

[jlevesy]

Hey @siyu6974, thanks for opening this issue.

In order to achieve load balancing, docker swarm allocates a virtual IP per service, and this VIP is load balanced between all known backends.

With the traefik.backend.loadbalancer.swarm=true, traefik will use this VIP instead of the containers IPs on the backend. From the traefik point of view, we're seeing only one IP, so having one backend is the correct behaviour.

We're going to clarify this on the documentation, thanks again for reporting.