jakarta-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: svn commit: r1164186 - in /jakarta/jmeter/trunk: bin/jmeter.properties src/protocol/http/org/apache/jmeter/protocol/http/sampler/HTTPHC4Impl.java xdocs/changes.xml xdocs/usermanual/component_reference.xml
Date Fri, 23 Sep 2011 15:36:45 GMT
On 23 September 2011 16:28, Milamber <milamber@apache.org> wrote:
>
>
> Le 23/09/2011 14:34, sebb a ecrit :
>> On 23 September 2011 15:20, Milamber <milamber@apache.org> wrote:
>>
>>> Hello,
>>>
>>> I've relaunched some tests with a new SimpleTest.jmx version which
>>> generate an unique URL for each request (in attachment)
>>> I did a analysis with wireshark to find the "retries GET" (same unique
>>> url call 2 times)
>>>
>>> NB "retries GET" means: GET /unique_url at end of TCP conversations
>>> without response, and a new retry GET /unique_url at the begin of TCP
>>> conversations
>>>
>>> Results (with 2.5.1RC1):
>>>
>>> a) Java and HC3.1 (and HC4) have "retries GET"
>>>
>>> b) The load test make ~15745 requests.
>>>
>>> * With Java : 670 "retries GET"
>>>
>>> * With HC3.1 (retrycount default) : 536 "retries GET"
>>>
>>> * With HC4 (retrycount force to 3) : 571 "retries GET"
>>>
>> Were any requests retried *more* than once in any of the implementations?
>> It would be interesting to know if the Java implementation can retry
>> more than once.
>>
>
> 0 request retried more than once for JAVA / HC 3 / HC 4
>
>
>>
>>> I suppose that the retrycount default value to 3 (or 1) is needed for HC4.
>>>
>> For backwards compatibility (and compatibility with HC3), then we
>> should choose 3.
>> I asked on the HC mailing list, and the choice of 3 was arbitrary.
>>
>> For compatibility with Java we should probably set them both to 1.
>>
>> I think 1 might be the best compromise - should catch most errors.
>>
>
> 1 retry count seems the good value from RFC HTTP/1.1 Keep-alive :
> "Clients which assume persistent connections and pipeline immediately
> after connection establishment SHOULD be prepared to retry their
> connection* if the first pipelined attempt fails*. If a client does such
> a retry, it MUST NOT pipeline before it knows the connection is
> persistent. Clients MUST also be prepared to resend their requests if
> the server closes the connection before sending all of the corresponding
> responses. "
>
> http://www.w3.org/Protocols/rfc2616/rfc2616-sec8.html#sec8.1.2.2

OK, thanks very much for investigating that.

Default retry count will be set to 1 for HC3 and HC4.

> Milamber
>
>> In any case, we need to document the settings.
>> I'll add some text to component reference shortly.
>>
>>
>>> Milamber
>>>
>>>
>>>
>>> Le 23/09/2011 12:58, sebb a ecrit :
>>>
>>>> Has anyone seen the errors with the Java implementation?
>>>>
>>>> There does not seem to be any documentation [1] on how many retries it
>>>> does, but I have seen something that suggests it normally retries just
>>>> once.
>>>>
>>>> [1] http://download.oracle.com/javase/6/docs/technotes/guides/net/properties.html
>>>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@jakarta.apache.org
>> For additional commands, e-mail: dev-help@jakarta.apache.org
>>
>>
>>
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: dev-help@jakarta.apache.org


Mime
View raw message