User's actions getting switched during load testing

[Matt Woodworth <ma...@cisco.com>]

Friends,

We're just about finished building a semi-large web app (~500 classes) 
but we're having problems during our load test.  The load tests have 
scripts (we're using load runner) that perform a series of actions 
(screen loads, posts, etc.) for a given user.  When he's finished doing 
his thing, he stops.  Different groups of users do different things.  A 
user from one group would be unable to perform an action that another 
group can perform.

The trouble is, we're seeing that users in group A are attempting to 
perform actions that only group b can do.  There is *no* chance that 
it's a problem is the load runner script.  They are fairly 
straightforward to check.  This only happens under load.  We can have 10 
users on the system for hours and hours and never see a problem.  If we 
raise that to 30 users we'll get the problem in a minute or two.

Our current theory is that somehow, sessions are getting crossed/mixed 
or whatever between users.  Turbine is performing an action for me when 
it was you who requested it.  This ties into another problem we're 
having.  If two users from group b perform the same action, sometimes a 
user gets the data from a different user.  This seems to be a different 
symptom of the same problem.  User A gets the results of user B's request.

I have another hint.  Increasing the services.PoolService.pool.capacity 
setting from 128 to 512 seems to have fixed the problem (or at least 
moved it up to beyond 70+ concurrent users, the current stopping place 
for our load test).

Does anybody know how to actually cure this problem, instead of simply 
pushing it up like I've done so far?

Matt Woodworth

PPS  I had orginally sent this to the wrong mailing list.  If you also 
noticed this question going to turbine-dev please don't think of me as a 
spammer or wicked cross poster.  It was just a mistake.


--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: User's actions getting switched during load testing

[Scott Eade <se...@backstagetech.com.au>]

Just in case it is relevant, over on the jetty-support mailing list they are
attempting to track down something like what you are experiencing.  Do you
happen to be using Jetty as your servlet container?

Scott
-- 
Scott Eade
Backstage Technologies Pty. Ltd.
http://www.backstagetech.com.au
.Mac Chat/AIM: seade at mac dot com


On 19/01/2003 3:03 AM, "Matt Woodworth" <ma...@cisco.com> wrote:

> Friends,
> 
> We're just about finished building a semi-large web app (~500 classes)
> but we're having problems during our load test.  The load tests have
> scripts (we're using load runner) that perform a series of actions
> (screen loads, posts, etc.) for a given user.  When he's finished doing
> his thing, he stops.  Different groups of users do different things.  A
> user from one group would be unable to perform an action that another
> group can perform.
> 
> The trouble is, we're seeing that users in group A are attempting to
> perform actions that only group b can do.  There is *no* chance that
> it's a problem is the load runner script.  They are fairly
> straightforward to check.  This only happens under load.  We can have 10
> users on the system for hours and hours and never see a problem.  If we
> raise that to 30 users we'll get the problem in a minute or two.
> 
> Our current theory is that somehow, sessions are getting crossed/mixed
> or whatever between users.  Turbine is performing an action for me when
> it was you who requested it.  This ties into another problem we're
> having.  If two users from group b perform the same action, sometimes a
> user gets the data from a different user.  This seems to be a different
> symptom of the same problem.  User A gets the results of user B's request.
> 
> I have another hint.  Increasing the services.PoolService.pool.capacity
> setting from 128 to 512 seems to have fixed the problem (or at least
> moved it up to beyond 70+ concurrent users, the current stopping place
> for our load test).
> 
> Does anybody know how to actually cure this problem, instead of simply
> pushing it up like I've done so far?
> 
> Matt Woodworth
> 
> PPS  I had orginally sent this to the wrong mailing list.  If you also
> noticed this question going to turbine-dev please don't think of me as a
> spammer or wicked cross poster.  It was just a mistake.
> 
> 
> --
> To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
> 


--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>