httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gregg L. Smith" <...@gknw.net>
Subject Re: Time for 2.4.0 GA??
Date Fri, 13 Jan 2012 02:06:13 GMT
On 1/12/2012 5:29 PM, Daniel Ruggeri wrote:
> On 1/12/2012 5:50 PM, Gregg L. Smith wrote:
>> Either apachehaus.com or apachelounge.com have 2.3.16 binaries
>> available for Windows.
>>
>> The problem is with the directive;
>> AcceptFilter httpd none
>>
>> That is the only non-stardard config option.
> Greg;
>     Thanks for the overview - if I understand correctly this seems to
> manifest with socket reuse and incomplete or empty responses - correct?
> And that this is a matter of recycling existing listening sockets rather
> than anything keepalive-related. If so, I'm guessing a simple perl LWP
> script to make request after request for a known-size resource (and
> verifying the size returned) would do. Any other pointers or things to
> look for? I seem to recall SSL being brought up, or is that a separate
> issue?
>
Hi Daniel,

I typo'd that again. I am bad at that, I should just copy and paste from 
my config.
AcceptFilter https none

SSL has everything to do with it. It is just in SSL with AcceptFilter 
for that protocol set to "none"

I find with trace logging on it doesn't want to manifest itself. When I 
turn off the trace logging, it happens often. It's odd, I do not get it.

Yes, incomplete or empty responses. The empty ones are easy to spot. 
Just an image here or there is a possible incomplete. It may be 
keepalive. On some windows flavors, we do not have a huge pool of 
connections and can be DOSed rather easy. I have seen some errors while 
trace logging was on, that may be suggesting that.

Looking, Keepalive is high, at 30, the default has been whittled down 
since I configured the server the 1st time (and have kept these config 
files) . Timeout is at 60. Even at 60,  I have seen this;

[Thu Jan 12 15:38:31.812658 2012] [ssl:info] [pid 3504:tid 720] (OS 
10060)A connection attempt failed because the connected party did not 
properly respond after a period of time, or established connection 
failed because connected host has failed to respond.  : [client 
192.168.1.1:51417] AH01991: SSL input filter read failed.

However, I can say this does not happen to just me, Steffen has seen 
same behavior.

It seems to happen more pronounced on my SNI hosts that the default SSL 
host. So maybe it is an artifact of SNI. A lot of possibilities I just 
cannot quite pin down.

II do not really have much problems with *not* setting AcceptFilter to 
none on https, not like I have on standard http where it has to be set 
to none or the server just stops responding after some time.




Mime
View raw message