You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by 1+ ∞ <wa...@foxmail.com> on 2019/02/15 02:03:57 UTC
回复： Jmeter 4.0 and jmeter 5.0 comparison
Thank you very much!
------------------ 原始邮件 ------------------
发件人: "Philippe Mouawad"<ph...@gmail.com>;
发送时间: 2019年2月15日(星期五) 凌晨5:00
收件人: "JMeter Users List"<us...@jmeter.apache.org>;
主题: Re: Jmeter 4.0 and jmeter 5.0 comparison
Hello,
There is a documented major change in JMeter 5.0 :
- https://jmeter.apache.org/changes.html#Incompatible%20changes
*Since JMeter 5.0, when using default HC4 Implementation, JMeter will reset
HTTP state (SSL State + Connections) on each thread group iteration. If you
don't want this behaviour, set
httpclient.reset_state_on_thread_group_iteration=false*
We do this, because a new Thread Group iteration is a new user in general,
so the SSL Handshake would occur again.
If a new Thread Group iteration is not a new user for you , set
httpclient.reset_state_on_thread_group_iteration=false
So first thing to do would be comparing same application version with
JMeter 4.0 and JMeter 5.0 with
httpclient.reset_state_on_thread_group_iteration=false.
If performances are similar, then either a new Thread Group iteration is a
new user and you need to set
httpclient.reset_state_on_thread_group_iteration=true
and then:
- Either your JMeter instance is overloaded, then increase its CPU
- Either it' s your application which is not able to handle the SSL
Handshake efficiently enough
Regards
On Wed, Feb 13, 2019 at 3:44 AM 1+ ∞ <wa...@foxmail.com> wrote:
> hi ，excuse me
>
>
> ------------------ 原始邮件 ------------------
> 发件人: "1+ ∞"<wa...@foxmail.com>;
> 发送时间: 2019年2月13日(星期三) 上午10:35
> 收件人: "user-subscribe"<us...@jmeter.apache.org>;
>
> 主题: Jmeter 4.0 and jmeter 5.0 comparison
>
>
>
> The same concurrency, jmeter4.0 and jmeter 5.0 running results in the tps
> and response time gap is very large, Jmeter 5.0 tps is more than 700, in
> the jmeter4.0 above tps reached more than 2,000。
>
>
> At the same time, when jmeter 5.0 is used, a large number of tcp states
> appear as time_wait connections.
>
>
> Note: The configuration of jmeter 5.0 and jmeter 4.0 is not only
> distributed, but the rest of the configuration is the default configuration
> after the official download.
--
Cordialement.
Philippe Mouawad.
Re: 回复： Jmeter 4.0 and jmeter 5.0 comparison
Posted by Felix Schumacher <fe...@internetallee.de>.
Am 15.02.19 um 07:46 schrieb 1+ ∞:
> In the case where jmeter 5.0 is
> httpclient.reset_state_on_thread_group_iteration=true, why is the
> close_wait state in the tcp connection a lot?
If my google foo didn't let me down, this indicates that the client
didn't close the connection yet. Can you post a minimal test plan that
reproduces that issue, that we could use?
Some ideas to investigate on your side. Do you use keepalive on your
requests? Is the behavior different, when you disable keepalive?
>
>
> httpclient.reset_state_on_thread_group_iteration=true
The mailing list strips most images from the mails (as it has done in
this case), so we can't see, what you want to show us.
Regards,
Felix
>
> httpclient.reset_state_on_thread_group_iteration=false
>
>
> ------------------ 原始邮件 ------------------
> *发件人:* "1+ ∞"<wa...@foxmail.com>;
> *发送时间:* 2019年2月15日(星期五) 上午10:03
> *收件人:* "JMeter Users List"<us...@jmeter.apache.org>;
> *主题:* 回复： Jmeter 4.0 and jmeter 5.0 comparison
>
> Thank you very much!
>
>
> ------------------ 原始邮件 ------------------
> 发件人: "Philippe Mouawad"<ph...@gmail.com>;
> 发送时间: 2019年2月15日(星期五) 凌晨5:00
> 收件人: "JMeter Users List"<us...@jmeter.apache.org>;
>
> 主题: Re: Jmeter 4.0 and jmeter 5.0 comparison
>
>
>
> Hello,
> There is a documented major change in JMeter 5.0 :
>
> - https://jmeter.apache.org/changes.html#Incompatible%20changes
>
> *Since JMeter 5.0, when using default HC4 Implementation, JMeter will
> reset
> HTTP state (SSL State + Connections) on each thread group iteration.
> If you
> don't want this behaviour, set
> httpclient.reset_state_on_thread_group_iteration=false*
>
> We do this, because a new Thread Group iteration is a new user in general,
> so the SSL Handshake would occur again.
> If a new Thread Group iteration is not a new user for you , set
> httpclient.reset_state_on_thread_group_iteration=false
>
> So first thing to do would be comparing same application version with
> JMeter 4.0 and JMeter 5.0 with
> httpclient.reset_state_on_thread_group_iteration=false.
>
> If performances are similar, then either a new Thread Group iteration is a
> new user and you need to set
> httpclient.reset_state_on_thread_group_iteration=true
> and then:
>
> - Either your JMeter instance is overloaded, then increase its CPU
> - Either it' s your application which is not able to handle the SSL
> Handshake efficiently enough
>
>
> Regards
>
> On Wed, Feb 13, 2019 at 3:44 AM 1+ ∞ <wa...@foxmail.com>
> wrote:
>
> > hi ，excuse me
> >
> >
> > ------------------ 原始邮件 ------------------
> > 发件人: "1+ ∞"<wa...@foxmail.com>;
> > 发送时间: 2019年2月13日(星期三) 上午10:35
> > 收件人: "user-subscribe"<us...@jmeter.apache.org>;
> >
> > 主题: Jmeter 4.0 and jmeter 5.0 comparison
> >
> >
> >
> > The same concurrency, jmeter4.0 and jmeter 5.0 running results in
> the tps
> > and response time gap is very large, Jmeter 5.0 tps is more than 700, in
> > the jmeter4.0 above tps reached more than 2,000。
> >
> >
> > At the same time, when jmeter 5.0 is used, a large number of tcp states
> > appear as time_wait connections.
> >
> >
> > Note: The configuration of jmeter 5.0 and jmeter 4.0 is not only
> > distributed, but the rest of the configuration is the default
> configuration
> > after the official download.
>
>
>
> --
> Cordialement.
> Philippe Mouawad.
回复： Jmeter 4.0 and jmeter 5.0 comparison
Posted by 1+ ∞ <wa...@foxmail.com>.
In the case where jmeter 5.0 is httpclient.reset_state_on_thread_group_iteration=true, why is the close_wait state in the tcp connection a lot?
httpclient.reset_state_on_thread_group_iteration=true
httpclient.reset_state_on_thread_group_iteration=false
------------------ 原始邮件 ------------------
发件人: "1+ ∞"<wa...@foxmail.com>;
发送时间: 2019年2月15日(星期五) 上午10:03
收件人: "JMeter Users List"<us...@jmeter.apache.org>;
主题: 回复： Jmeter 4.0 and jmeter 5.0 comparison
Thank you very much!
------------------ 原始邮件 ------------------
发件人: "Philippe Mouawad"<ph...@gmail.com>;
发送时间: 2019年2月15日(星期五) 凌晨5:00
收件人: "JMeter Users List"<us...@jmeter.apache.org>;
主题: Re: Jmeter 4.0 and jmeter 5.0 comparison
Hello,
There is a documented major change in JMeter 5.0 :
- https://jmeter.apache.org/changes.html#Incompatible%20changes
*Since JMeter 5.0, when using default HC4 Implementation, JMeter will reset
HTTP state (SSL State + Connections) on each thread group iteration. If you
don't want this behaviour, set
httpclient.reset_state_on_thread_group_iteration=false*
We do this, because a new Thread Group iteration is a new user in general,
so the SSL Handshake would occur again.
If a new Thread Group iteration is not a new user for you , set
httpclient.reset_state_on_thread_group_iteration=false
So first thing to do would be comparing same application version with
JMeter 4.0 and JMeter 5.0 with
httpclient.reset_state_on_thread_group_iteration=false.
If performances are similar, then either a new Thread Group iteration is a
new user and you need to set
httpclient.reset_state_on_thread_group_iteration=true
and then:
- Either your JMeter instance is overloaded, then increase its CPU
- Either it' s your application which is not able to handle the SSL
Handshake efficiently enough
Regards
On Wed, Feb 13, 2019 at 3:44 AM 1+ ∞ <wa...@foxmail.com> wrote:
> hi ，excuse me
>
>
> ------------------ 原始邮件 ------------------
> 发件人: "1+ ∞"<wa...@foxmail.com>;
> 发送时间: 2019年2月13日(星期三) 上午10:35
> 收件人: "user-subscribe"<us...@jmeter.apache.org>;
>
> 主题: Jmeter 4.0 and jmeter 5.0 comparison
>
>
>
> The same concurrency, jmeter4.0 and jmeter 5.0 running results in the tps
> and response time gap is very large, Jmeter 5.0 tps is more than 700, in
> the jmeter4.0 above tps reached more than 2,000。
>
>
> At the same time, when jmeter 5.0 is used, a large number of tcp states
> appear as time_wait connections.
>
>
> Note: The configuration of jmeter 5.0 and jmeter 4.0 is not only
> distributed, but the rest of the configuration is the default configuration
> after the official download.
--
Cordialement.
Philippe Mouawad.