I have upgraded to Servicemix 3.0 and load tested it again by deploying
Servicemix on Weblogic 8.1 and Geronimo 1.0 app servers. The situation
seems to be improved but the overall problem still persists.
For #4 in my mail below, servicemix 2.0.2 used to get stuck for around
30 transactions. But now for Servicemix 3.0 this runs well for around
1000 transactions.
I have debugged the code and found that the messages are being put in
the BoundedLinkedQueue but, the work is not getting scheduled in the
run() method of SedaQueue.java after around 1000 transactions.
For the flow JmsIn --> JmsOut, the put() method is called only once,
which is perfect, as the message exchange needs to flow thru the NMR
only once from JmsIn to JmsOut.
But, for the flow JmsIn --> WSIF --> JmsOut, ideally speaking the put()
method has to be invoked only twice. But, the method is invoked 4 times.
I checked the destination service from the message exchange in the put()
method of BoundedLinkedQueue.java and run() method of 'new Work()' inner
class. In both the methods, the destination service for the first two
times was as expected i.e. WSIF and the JmsOut. The flow should stop
here. But, put() method is called twice more, the destination service
being WSIF. And run() method called once more and the destionation
service being WSIF again. I think because of excessive puts in the
BoundedLinkedQueue, the scheduling of the work is getting affected.
The servicemix.xml being:
<sm:container spring:id="jbi"
useMBeanServer="false"
createMBeanServer="false"
dumpStats="true"
statsInterval="10">
<sm:activationSpecs>
<!-- Subscribe to a JMS destination -->
<sm:activationSpec componentName="inputReceiver"
service="my:inputReceiver"
destinationService="my:Inventory_EJBService">
<sm:component>
<bean xmlns="http://xbean.org/schemas/spring/1.0";
class="com.infosys.servicemix.components.jms.JmsInWLBinding">
<property
name="queueConnFactory" value="java:comp/env/infosysServicemixQCF"/>
<property name="qu"
value="java:comp/env/Queue1"/>
</bean>
</sm:component>
</sm:activationSpec>
<sm:activationSpec componentName="Inventory_EJBService"
service="my:Inventory_EJBService" destinationService="my:outputSender">
<sm:component>
<bean
xmlns="http://xbean.org/schemas/spring/1.0";
class="org.apache.servicemix.components.wsif.WSIFBinding">
<property name="definitionResource"
value="MultiAddition.wsdl"/>
</bean>
</sm:component>
</sm:activationSpec>
<sm:activationSpec componentName="outputSender"
service="my:outputSender">
<sm:component>
<bean xmlns="http://xbean.org/schemas/spring/1.0";
class="com.infosys.servicemix.components.jms.JmsOutWLBinding">
<property
name="queueConnFactory" value="java:comp/env/infosysServicemixQCF"/>
<property name="qu"
value="java:comp/env/Queue2"/>
</bean>
</sm:component>
</sm:activationSpec>
Thanks n Regards,
Ujval Mysore
-----Original Message-----
From: Guillaume Nodet [mailto:[EMAIL PROTECTED]
Sent: Monday, August 28, 2006 6:26 PM
To: [email protected]
Subject: Re: Issues found when load tested servicemix
30 is the default size of ServiceMix thread pool.
So it means a component does not correctly implement the MEP used.
>From your explanations, it should be the jms out binding.
What's the configuration you used for #4 ?
Also, you should try to upgrade to 3.0-M2, it will certainly fix the
problem.
On 8/28/06, Ujval Mysore <[EMAIL PROTECTED]> wrote:
>
>
> Hi,
>
>
>
> Right now I am evaluating servicemix 2.0.2. As part of other
> requirements I have deployed servicemix2.0.2 on Weblogic 8.1. I have
> load tested servicemix in various scenarios and I was confused to see
> the results.
>
>
>
> I have load tested all the following scenarios using load runner. I
ran
> the load tests for 5 minutes with 15 simultaneous virtual users, each
> vuser resending the request once it gets back the response for its
> previous request and with a very small payload of 10KB per request.
>
>
>
> Various scenarios include orchestrating the binding components JMS,
> router and WSIF in the following ways:
>
>
>
> 1) JMSInBinding --> JMSOutBinding:
>
> The client sends the request to servicemix. The JMSInBinding which is
> listening to a queue reads the message and sends the normalized
message
> to the JMSOutBinding which simply creates a jms message and puts it in
> another queue.
>
> This worked fine without any problem.
>
>
>
> 2) JMSInBinding --> WSIF:
>
> The JMSInBinding sends the message to the WSIF binding component. This
> scenario also worked fine.
>
>
>
> 3) JMSInBinding --> WSIF --> WSIF
>
> Adding to the previous scenario, the destination service of the
> WSIF will be one more WSIF. This scenario also worked fine.
>
>
>
> 4) JMSInBinding --> WSIF --> JMSOutBinding
>
> This scenario failed miserably. It hardly processes 30
> requests and then hangs. I have debugged the servicemix code and found
> that for the 30 successful requests, all the three components executed
> successfully. For the failed transactions, the flow stops at the
> JMSInBinding. It seems a deadlock is occurring somewhere.
>
>
>
> On further debugging I found that the method
doSend(MessageExchangeImpl)
> in the SedaFlow.java is also getting executed. When the request flows
> from JMSInBinding to the WSIF, the values me.getSyncState() and
> me.getMirror().getSyncState() are '0'. Hence enqueue(packet) is being
> called. After the execution of WSIF before going to JMSOutBinding, I
> found the values to be me.getSyncState() =0 and
> me.getMirror().getSyncState() = 1. Hence doRouting is getting called.
>
> This is getting executed perfectly for around first 30 requests.
Later,
> all the requests are getting stuck in the first flow i.e. JMSInBinding
> --> WSIF. The WSIF is also not getting executed.
>
>
>
> I also found out that queue.poll method in run() of SedaQueue.java is
> working properly, as it is getting the message exchange without any
> problem.
>
>
>
> I think there is a deadlock occurring somewhere here.
>
>
>
> 5) JMSInBinding --> Router --> JMSOutBinding
>
> I was surprised to note that this scenario works without any problem.
>
>
>
> 6) JMSInBinding --> Router --> WSIF
>
> This scenario too worked fine.
>
>
>
> 7) JMSInBinding --> Router --> WSIF --> JMSOutBinding
>
> This scenario also failed. The results were similar to the point 4.
>
>
>
> Is it a problem with servicemix? Kindly look into this and please let
me
> know at the earliest.
>
>
>
> Is there any mechanism to trace the flow of the message in servicemix?
>
>
>
> Thanks in advance,
>
> Ujval Mysore
>
>
>
>
>
> **************** CAUTION - Disclaimer *****************
> This e-mail contains PRIVILEGED AND CONFIDENTIAL INFORMATION intended
> solely for the use of the addressee(s). If you are not the intended
> recipient, please notify the sender by e-mail and delete the original
> message. Further, you are not to copy, disclose, or distribute this
e-mail
> or its contents to any other person and any such actions are unlawful.
This
> e-mail may contain viruses. Infosys has taken every reasonable
precaution to
> minimize this risk, but is not liable for any damage you may sustain
as a
> result of any virus in this e-mail. You should carry out your own
virus
> checks before opening the e-mail or attachment. Infosys reserves the
right
> to monitor and review the content of all messages sent to or from this
> e-mail address. Messages sent to or from this e-mail address may be
stored
> on the Infosys e-mail system.
> ***INFOSYS******** End of Disclaimer ********INFOSYS***
>
--
Cheers,
Guillaume Nodet
