Mailing-List: contact user-help@jmeter.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "JMeter Users List" <user@jmeter.apache.org>
MIME-Version: 1.0
In-Reply-To: <CAEvnqUia9bGdOBp+DaRdvb7qTLXh6PRs5xBKzfsg421Co-dTdQ@mail.gmail.com>
References: <CAEvnqUia9bGdOBp+DaRdvb7qTLXh6PRs5xBKzfsg421Co-dTdQ@mail.gmail.com>
From: Deepak Shetty <shettyd@gmail.com>
Date: Wed, 27 Dec 2017 13:49:02 -0800
Message-ID: <CAKUYAYeMDf3yywwCYgZo3VuCwdJV-72jP9GScBLx6+X=_5Vf_w@mail.gmail.com>
Subject: Re: Gateway Timeout Errors
To: JMeter Users List <user@jmeter.apache.org>
Content-Type: multipart/alternative; boundary="94eb2c0328bc1f01ea0561595e03"
archived-at: Wed, 27 Dec 2017 21:49:13 -0000

--94eb2c0328bc1f01ea0561595e03
Content-Type: text/plain; charset="UTF-8"

>What would be the reason for this
Difficult to say , its the app thats behaving badly!. What are the response
times you are seeing ? An example of systems that exhibit the behavior you
describe are systems that lazily load caches (with synchronized wait
behavior for other requests while the cache is being loaded OR all of the
other requests also try to load the cache) and hence initial requests take
a long time and so for e.g. if your first request is going to take 20
seconds then you will end up with 20 requests waiting (50 users with ramp
up of 50 seconds is 1 new request every second). If all are trying to load
cache you might even see more as everyone slows down everyone else.
By increasing your ramp-up you are reducing the initial load the system
sees - But does each thread just run 1 iteration only ? In which case yes ,
you probably have very little concurrency.

Note , for most test scripts , Ramp Up is meant to prevent a burst of
requests which may not represent your normal load. Your concurrency
*should* be determined by other factors (the duration your test runs after
all threads are created , the think times in your script etc). Think about
it this way. Say 50 users are sitting doing something on your website - not
everyone is actually doing something at the same time - most users will be
reading the page for e.g. - but maybe 10 of those click a link at the same
time - the concurrency your system sees is approx 10. In your test you
might still model this as 50 threads (for 50 users) but create your script
such that the system sees about 10 requests at the same time. What the Ramp
up controls is that your system doesnt see 50 requests immediately as soon
as the test starts .


On Tue, Dec 19, 2017 at 12:44 AM, Riji K <krijiram@gmail.com> wrote:

> Hi All,
>
> I'm running a 50 User concurrent test with 50seconds RUT on a
> web-application, which results in higher number of "Gateway Timeout
> Errors".
>
> With Increasing the ramp-up to 1000 seconds RUT, the errors were not
> observed.
>
> What would be the reason for this. I think 50 seconds  ramp-up is
> sufficient for the user. Also would i like to know if i give 1000 seconds
> ramp-up for 50 user, will be any concurrency.?
>
> Please help me.
>
> Regards,
> Riji K
>

--94eb2c0328bc1f01ea0561595e03--