rancher.metadata container name collision in backend (different behaviour compared to rancher.api)

[PicoCreator]

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

Bug (Unless its intentional)

What did you do?

Docker System : Rancher v1.6.5
Deployed latest official image : traefik:v1.4.0-rc2-alpine

With the following command for meta-data mode (formatted to be easier to read)

--web --rancher 
	--rancher.domain=rancher.localhost 
	--rancher.enableservicehealthfilter=false 
	--rancher.exposedbydefault=false 
	--rancher.metadata

Along with the following container labels

io.rancher.container.agent.role=environment
io.rancher.container.create_agent=true

Multiple stacks was then created, with different stack names. But same container names (web-studio in this case)

Each container is defined with the following traefik labels

traefik.enable=true
traefik.frontend.rule=Host:the-container-unique-url.domain.something
traefik.port=80

Note: this is a followup from #2125

What did you expect to see?

Every container, from each stack, with its own "host name rule" for frontend, and its own backend, as seen below, when traefik is using rancher.api

What did you see instead?

A single backend, for all the various frontend, when traefik is using rancher.metadata

Note : that both this traefik configuration was deployed side by side to the same rancher testing infrastructure. So both configuration was expected to be same

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

Version:      v1.4.0-rc2
Codename:     roquefort
Go version:   go1.9
Built:        2017-09-08_07:31:37PM
OS/Arch:      linux/amd64

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

None: See setup above

[SantoDE]

Hi @PicoCreator ,

this is already fixed in #2107, so it should be available with the next rc for 1.4 :)