Support for reading Apache Aurora serverssets from Zookeeper about traefik HOT 11 OPEN

I would like this as well.

from traefik.

Hi
Maybe It would be better to interact with Zookeeper directly instead no ?

from traefik.

Do you mean the clients connecting to the service should do the service lookup through ZooKeeper? If so, then that can be problematic since it pulls in ZooKeeper dependencies to all clients. Many of my clients today are just plain cURL/bash scripts that would be difficult to interface with ZK.

from traefik.

No, that's not what I wanted to say. Aurora does not provide any rest API to interact with. As Aurora puts its jobs in Zookeper, I was thinking of getting the conf directly from Zookeper :)

from traefik.

Right, Aurora will publish metadata for service discovery about every task it launches as a service in zookeeper. What we are asking for is that traefik be taught to read that metadata and do load balancing / proxying to the instances.

from traefik.

OK, but I will need some help, or some time :)

from traefik.

This is the most commonly used library for go + zk. I could write a small python script to create the "server set" znodes for you if that would be a good starting point?

from traefik.

#25 is relevant somewhat as libkv supports zookeeper.

from traefik.

@SEJeff You're absolutely right :) But it will need some adaptations to support Apache Aurora standard. It will not run out of the box.

from traefik.

@emilevauge Now the serverset data would be written by Aurora and should be read-only as far as traefik is concerned. Would it make more sense to just build support for it in natively?

from traefik.

Hi, as Zookeeper is now supported, it should be quite easy to get Aurora on the same way :) Help is welcomed :)

from traefik.