Add current_queue metric for servers

[discordianfish]

Test it here and looks good.

[grobie]

It's not queueing servers but requests. Maybe let's improve this and the backend's help message. First idea:

server: Current number of queued requests assigned to this server.
backend: Current number of queued requests not assigned to any server.

[grobie]

👍 after improving the help messages.

[discordianfish]

@grobie I've changed the description but are we sure that it's correct? I would have assumed that backend queue is the sum of all servers queue, no the spillover where it can't assign requests to any servers anymore..?

[discordianfish]

Reading the haproxy docs I guess you're right, although I'm not completely sure..
If you are, feel free to merge this PR.

[grobie]

Let's adjust the max queue help as well for consistency.

[grobie]

I haven't come across a HAProxy setup which is configured to queue in a while. At SoundCloud none of our servers have a configured maxconn, therefore HAProxy will never queue any requests.

I can't tell you when it's possible that a request is queued on a server, and not the global queue. In our old mothership routing stacks I believe only the global queue was used.