Rainbows! Rack HTTP server user/dev discussion
 help / color / mirror / code / Atom feed
From: Eric Wong <normalperson-rMlxZR9MS24@public.gmane.org>
To: Rainbows! list <rainbows-talk-GrnCvJ7WPxnNLxjTenLetw@public.gmane.org>
Subject: Re: Transfer encoding chunked disables HAProxy gzip compression
Date: Fri, 2 Aug 2013 20:58:10 +0000	[thread overview]
Message-ID: <20130802205810.GA9486@dcvr.yhbt.net> (raw)
In-Reply-To: <CD8781D3-288B-4B61-85ED-16B8B15A90E9-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

Claudio Poli <claudio-3HQ/CcOImoi171pxa8y+qA@public.gmane.org> wrote:
> Hello, story time.
> 
> I was investigating why suddenly our service become slow in response
> times, currenrly the app is served through Rainbows! behind HAProxy.
> 
> It turns out that the problem is HAProxy, which we use as
> frontend/load balancer *and* for gzip compression of responses from
> backends.

<snip>
OK, hopefully these problems get fixed in HAProxy soon

> So, I tried to study rainbows and unicorn internals a bit (-N switch,
> -E 'development, deploy, none') to attempt and disable chunked
> middleware and replace with a Rack::ContentLength, so if and when the
> body can be enumerated my responses would have a fixed size length to
> feed the client.

> It doesn't work: I tried to place them in config.ru, in rails'
> application.rb, tried the -N switch, messed with -E but in staging
> every response that comes out are of type http/1.1 chunked requests.

This could be from Rails chunked responses, especially if you're using
-N already.  You may have to dig into Rails internals a bit...

When in doubt, try dropping Rails from the equation and using a bare
bones "hello world" type Rack app (e.g. Rack::Lobster)

Rack::ContentLength won't add a Content-Length header if the response
was already chunked anywhere before.

> So I went ahead and added Rack::Deflater to live with the chunked
> response types and enable gzip compression as well, even if the cpu
> spent time compressing moved from one load balancer to the app server.

This is probably the better way to go anyways, especially if you add
more Rainbows! workers and maintain few HAproxy processes.

> I would like to know more about this behavior and based on which
> factors Rainbows! or Unicorn set and pick the transfer-encoding or
> content-lenght to use.

It's down to middleware/Rack stack ordering (LIFO), and reading code of
these middlewares.  Since Rack::Chunked runs first in the default
Rack::Server configs, Chunked can insert its headers before
ContentLength for HTTP 1.1 requests.

Rack::Chunked won't force an existing Content-Length response to
be chunked, it'll only make things chunked that don't have either
Content-Length/Transfer-Encoding set.
_______________________________________________
Rainbows! mailing list - rainbows-talk-GrnCvJ7WPxnNLxjTenLetw@public.gmane.org
http://rubyforge.org/mailman/listinfo/rainbows-talk
Do not quote signatures (like this one) or top post when replying


  parent reply	other threads:[~2013-08-02 21:50 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-08-02 19:26 Transfer encoding chunked disables HAProxy gzip compression Claudio Poli
     [not found] ` <CD8781D3-288B-4B61-85ED-16B8B15A90E9-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-08-02 20:03   ` Claudio Poli
     [not found]     ` <9C3CB455-F4F3-4DF4-BF67-1801547FE1B7-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-08-02 20:46       ` Eric Wong
     [not found]         ` <20130802204647.GA7479-yBiyF41qdooeIZ0/mPfg9Q@public.gmane.org>
2013-08-02 21:00           ` Eric Wong
     [not found]             ` <20130802210004.GB9486-yBiyF41qdooeIZ0/mPfg9Q@public.gmane.org>
2013-08-02 21:35               ` Claudio Poli
2013-08-02 21:49               ` Claudio Poli
2013-08-02 20:58   ` Eric Wong [this message]
     [not found]     ` <20130802205810.GA9486-yBiyF41qdooeIZ0/mPfg9Q@public.gmane.org>
2013-08-02 22:34       ` Claudio Poli
     [not found]         ` <FE4BFDCF-538B-4D2D-8991-FE0047DF52AA-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-08-02 23:11           ` Eric Wong
     [not found]             ` <20130802231115.GA24708-yBiyF41qdooeIZ0/mPfg9Q@public.gmane.org>
2013-08-03  0:20               ` Claudio Poli
     [not found]                 ` <E85CF5AB-7FF6-4BE8-BAAC-5416E14D4A14-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-08-03  1:06                   ` Eric Wong
     [not found]                     ` <20130803010627.GA4521-yBiyF41qdooeIZ0/mPfg9Q@public.gmane.org>
2013-08-03  1:53                       ` Claudio Poli
     [not found]                         ` <9580D763-16FA-4DB5-8CA7-0648DD6788CD-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2013-08-03  2:30                           ` Eric Wong

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://yhbt.net/rainbows/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130802205810.GA9486@dcvr.yhbt.net \
    --to=normalperson-rmlxzr9ms24@public.gmane.org \
    --cc=rainbows-talk-GrnCvJ7WPxnNLxjTenLetw@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://yhbt.net/rainbows.git/

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).