Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 23 Apr 2008 14:23:55 +0800
From:      "Adrian Chadd" <adrian@freebsd.org>
To:        "Jeremy Chadwick" <koitsu@freebsd.org>
Cc:        freebsd-stable@freebsd.org, Ivan Voras <ivoras@freebsd.org>
Subject:   Re: ab2 (apache benchmark) problem
Message-ID:  <d763ac660804222323mc0e156ak6234757d72923b8e@mail.gmail.com>
In-Reply-To: <20080422195112.GA45977@eos.sc1.parodius.com>
References:  <fuivn7$9et$1@ger.gmane.org> <20080422195112.GA45977@eos.sc1.parodius.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On 23/04/2008, Jeremy Chadwick <koitsu@freebsd.org> wrote:
> On Mon, Apr 21, 2008 at 11:07:43PM +0200, Ivan Voras wrote:
>  > Every now and then people are complaining about the bug in ab2 that makes
>  > it unusable for benchmarking from FreeBSD (as a client). "ab2" is a HTTP
>  > benchmark that's bundled with the Apache web server. I found the apparent
>  > solution and I'd like to invite everyone who's interested to register on
>  > Apache's bugzilla and vote for the bug to show their interest
>  > (https://issues.apache.org/bugzilla/show_bug.cgi?id=44584).
>
>
> Well that's just peachy.  For years I'd been trying to figure out why ab
>  would work fine with a very small number of requests, but would block
>  indefinitely when performing large sums of them.  I'll be sure to vote
>  that up.

If its anything like ab1, the freaking async connect() call doesn't
even check the socket status after the async call returns.

I've got a locally modified apachebench1 which uses libevent and has
some tidyups applied - it isn't multithreaded, but it can do ~ gigabit
of testing traffic on a single CPU..



adrian


-- 
Adrian Chadd - adrian@freebsd.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?d763ac660804222323mc0e156ak6234757d72923b8e>