Re: check_client_connection using getsockopt(2)

[Simon Eskildsen <simon.eskildsen@shopify.com>]

On Wed, Feb 22, 2017 at 8:42 PM, Eric Wong <e@80x24.org> wrote:
> Simon Eskildsen <simon.eskildsen@shopify.com> wrote:
>
> <snip> Thanks for the writeup!
>
> Another sidenote: It seems nginx <-> unicorn is a bit odd
> for deployment in a containerized environment(*).
>
>> I meant to ask, in Raindrops why do you use the netlink API to get the
>> socket backlog instead of `getsockopt(2)` with `TCP_INFO` to get
>> `tcpi_unacked`? (as described in
>> http://www.ryanfrantz.com/posts/apache-tcp-backlog/) We use this to
>> monitor socket backlogs with a sidekick Ruby daemon. Although we're
>> looking to replace it with a middleware to simplify for Kubernetes.
>> It's one of our main metrics for monitoring performance, especially
>> around deploys.
>
> The netlink API allows independently-spawned processes to
> monitor others; so it can be system-wide.  TCP_INFO requires the
> process doing the checking to also have the socket open.
>
> I guess this reasoning for using netlink is invalid for containers,
> though...

If you namespace the network it's problematic, yeah. I'm considering
right now putting Raindrops in a middleware with the netlink API
inside the container, but it feels weird. That said, if you consider
the alternative of using `getsockopt(2)` on the listening socket, I
don't know how you'd get access to the Unicorn listening socket from a
middleware. Would it be nuts to expose a hook in Unicorn that allows
periodic execution for monitoring listening stats from Raindrops on
the listening socket? It seems somewhat of a narrow use-case, but on
the other hand I'm also not a fan of doing
`Raindrops::Linux.tcp_listener_stats("localhost:#{ENV["PORT"}")`, but
that might be less ugly.

>
>> I was going to use `env["unicorn.socket"]`/`env["puma.socket"]`, but
>> you could also do `env.delete("hijack_io")` after hijacking to allow
>> Unicorn to still render the response. Unfortunately the
>> `<webserver>.socket` key is not part of the Rack standard, so I'm
>> hesitant to use it. When this gets into Unicorn I'm planning to
>> propose it upstream to Puma as well.
>
> I was going to say env.delete("rack.hijack_io") is dangerous
> (since env could be copied by middleware); but apparently not:
> rack.hijack won't work with a copied env, either.
> You only need to delete with the same env object you call
> rack.hijack with.
>
> But calling rack.hijack followed by env.delete may still
> have unintended side-effects in other servers; so I guess
> we (again) cannot rely on hijack working portably.

Exactly, it gives the illusion of portability but e.g. Puma stores an
instance variable to check whether a middleware hijacked, rendering
the `env#delete` option useless.

>
>> Cool. How would you suggest I check for TCP_INFO compatible platforms
>> in Unicorn? Is `RUBY_PLATFORM.ends_with?("linux".freeze)` sufficient
>> or do you prefer another mechanism? I agree that we should fall back
>> to the write hack on other platforms.
>
> The Raindrops::TCP_Info class should be undefined on unsupported
> platforms, so I think you can check for that, instead.  Then it
> should be transparent to add FreeBSD support from unicorn's
> perspective.

Perfect. I'll start working on a patch.

>
>
> (*) So I've been wondering if adding a "unicorn-mode" to an
>     existing C10K, slow-client-capable Ruby/Rack server +
>     reverse proxy makes sense for containerized deploys.
>     Maybe...

I'd love to hear more about this idea. What are you contemplating?