Load Test Numbers with Different idp.properties settings

[Nate Klingenstein]

Yeah, if I drop it to 1 process and 1 thread per Grinder node, the TPS plunges but I get no errors.

On Jul 8, 2014, at 11:04 AM, Nate Klingenstein <ndk at internet2.edu>
 wrote:

> The error rate, from what I can tell, is an artifact of concurrency issues in the Grinder script/framework rather than anything in the IdP.  I think the IdP is clean.
> 
> 7/8/14 5:01:30 PM (thread 1 run 401 test 2): Aborted run due to Jython exception: NameError: wssorq [calling TestRunner]
> NameError: wssorq
>        File "/opt/grinder-framework/tests/saml2sso_idp/test.py", line 255, in __call__
> 
> 7/8/14 5:01:30 PM (thread 3 run 397 test 2): Success!
> 7/8/14 5:01:30 PM (thread 0 run 401 test 2): Success!
> 7/8/14 5:01:30 PM (thread 1 run 401 test 2): Failure!
> 7/8/14 5:01:30 PM (thread 1 run 401 test 2): ERROR ("Aborted run due to J..."), see error log for details
> 7/8/14 5:01:30 PM (thread 2 run 407 test 2): Success!
> 7/8/14 5:01:30 PM (thread 4 run 405 test 2): Success!
> 
> Also, idp.sesson.StorageService was typoed and I pasted that verbatim.  Correcting it to idp.session.StorageService didn't result in any performance change either.
> 
> On Jul 8, 2014, at 10:56 AM, "Cantor, Scott" <cantor.2 at osu.edu>
> wrote:
> 
>> On 7/8/14, 12:34 PM, "Nate Klingenstein" <ndk at internet2.edu> wrote:
>> 
>>> I haven't gone back to track down what the errors are yet.  I'll do that
>>> next.  The groupings of properties and defaults wasn't obvious to me, so
>>> apologies for the useless data.
>>> 
>>> No obvious performance degradation from:
>>> 
>>> idp.transientId.generator = shibboleth.StoredTransientIdGenerator
>>> idp.sesson.StorageService = shibboleth.StorageService
>>> 
>>> 107, 95, .06%
>> 
>> No improvement either, I guess. That rules out the crypto overhead I was
>> thinking of. That's useful to know anyway.
>> 
>> -- Scott
>> 
>> --
>> To unsubscribe from this list send an email to dev-unsubscribe at shibboleth.net
> 
> --
> To unsubscribe from this list send an email to dev-unsubscribe at shibboleth.net