Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Date: Thu, 2 Feb 2017 13:31:25 +0100 (MET)
From: Niklas Edmundsson <nikke@acc.umu.se>
To: httpd-dev <dev@httpd.apache.org>
Subject: httpd 2.4.25, mpm_event, ssl: Status of async write completion?
Message-ID: <alpine.GSO.2.20.1702021314290.14372@hatchepsut.acc.umu.se>
User-Agent: Alpine 2.20 (GSO 67 2015-01-07)
MIME-Version: 1.0
Content-Type: text/plain; format=flowed; charset=US-ASCII
archived-at: Thu, 02 Feb 2017 12:32:04 -0000


Hi all!

As we're seeing more and more https on ftp.acc.umu.se I've noticed 
that the number of threads listed as state W in server-status has 
skyrocketed.

From 2-4 threads busy using http we're talking 70-100 threads for the 
same bandwidth when the machine is pushing a mighty 2.7% average CPU 
load.

This is on a large-file workload, serving plain files.

I anticipated a few more threads due to ssl, but for slow downloaders 
I would expect connections doing async write and mostly waiting for 
the bits to dribble through the Internet Tubes inbetween the 
occational wakeup to do ssl&send...

Are we not doing async write completion at all on https/ssl and 
falling back to the old worker behaviour of one thread per connection?


/Nikke
-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
  Niklas Edmundsson, Admin @ {acc,hpc2n}.umu.se      |     nikke@acc.umu.se
---------------------------------------------------------------------------
  At last I'm organized", he sighed, and died.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=