On Feb 13, 2009, at 3:32 PM, Willy Tarreau wrote:
> On Fri, Feb 13, 2009 at 09:30:18PM +0100, Willy Tarreau wrote:
>> On Wed, Feb 11, 2009 at 06:34:35PM -0500, Luke Melia wrote:
>>> Has anyone seen this type of http response while using HAProxy?
>>> http://skitch.com/lukemelia/brm8t/strange-response
>>>
>>> It only occurs rarely and only started after we added HAProxy to our
>>> environment. Our stack is nginx -> haproxy -> mongrels.
>>>
>>> Other than this rare issue, HAProxy has been great (and I'm not 100%
>>> sure this is an HAProxy issue). Thanks much to Willy and the the
>>> other
>>> contributors!'
>>
>> Wow. That's pretty strange. I don't see any possibility for haproxy
>> to
>> do something like this, especially at the beginning of the data.
>> But I
>> can't imagine how nor why nginx would do that either. Also, it cannot
>> come from the mongrel, otherwise haproxy would have returned a 502
>> error
>> instead. However, I see that nginx replies "keep-alive". Would it be
>> possible that the request was the second one in a row and that there
>> is a problem somewhere with the previous response not been completely
>> consumed by the client ? For this, it would help if you could try
>> with
>> another client then try to disable keep-alive on either your browser
>> or nginx.
>
> and that's even more likely since the response has been compressed !
I can't reliably replicate this, unfortunately. I have however, seen screenshots from multiple clients (Firefox & IE on different machines), so I don't think it's client dependent. I'll look into the keep-alive idea.
A colleague I showed this to commented that he often used to see this behavior running FastCGI with Apache and mongrel behind. His explanation was for that was something about a "dirty buffer". Not sure if that's any help.
-- Luke Melia luke#lukemelia.com http://www.lukemelia.com/Received on 2009/02/13 21:48
This archive was generated by hypermail 2.2.0 : 2009/02/13 23:00 CET