httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dietmar Müller <oas.pub...@gmail.com>
Subject Re: [users@httpd] Re: large amount of R in server-status
Date Fri, 12 Mar 2010 10:16:30 GMT
2010/3/12 Hendrik Schmieder <hendrik.schmieder@jedox.com>

> Dietmar Müller schrieb:
>
>   Hi,
>>>
>>>  I use since long time the apache in version 2.0 on windows 2000 (SP4).
>>>  Since 4 weeks I can see (some times per day) many R in server-status.
>>>  Sometimes so much that the server run full and stop working because all
>>>
>>
>>   threads are used. I search in web and generally the response is that it
>>> have
>>>  to do with dos but I'm not sure (or can't believe that)
>>>
>>>  When I watch over some time the serverstatus I can see sometimes
>>> requests
>>>
>>
>>   staying in R status until the configparameter timeout=300 (default) kick
>>> the
>>>  request out. Sometimes this R staying requests use more then 50% from my
>>>  reads and then after timeout all works like expexcted. Sometimes the
>>>
>>
>>   complete webserver is blocked.
>>>
>>>  Current I have 450 threads on apache and in background 2 clustered
>>> tomcats
>>>  with a 200 threads.
>>>
>>>  Can anybody explain me how it is possible that some (?) can send a rq to
>>> a
>>>
>>
>>   webserver and don't finish the rq that the server wait for it? How can I
>>>  solve such situations?
>>>
>>>  Further: how can I identify and/or analyze a problem like this. I have
>>> now a
>>>  situation where I have log the network trafik with wireshark.
>>>
>>
>>
>>>  This problem start (about) at a time when we switch to a new provider.
>>>
>>>  I have now version 2.0.57 but I have the problem also in the latest
>>> version
>>>  from the 2.0 tree.
>>>
>>>  If someone has some hint and or experience how to identify the problem
>>> -->
>>>
>>
>>   thanks for any comment / hint
>>>
>>>  regards Dietmar
>>>
>>
>> Hi,
>> I found the problem :-)
>>
>> http://ha.ckers.org/slowloris
>>
>> /* for the archive */
>> to isolate the problem I write a small script which store every sec
>> netstat -na
>>
>> and the after analyze it and search some hours in web I found the problem.
>>
>> Might after migration to an other os follow mod is a solution:
>> http://mod-qos.sourceforge.net/
>>
>>
>> hope this help other people!
>>
>> If someone have any input about this pls let me know
>>
>> regards Dietmar
>>
>>
> And what is the root of your problem ?
>
> regards
>
>  Hendrik
>
>
The root problem is, that the apache webserver is susceptible for a small
perl script what send incomplete / slow http requests and this cause that
all threads from the httpd are in reading state.

Dietmar

Mime
View raw message