camel-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raul Kripalani (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CAMEL-5390) Option to assign unique correlation ID to JMS messages
Date Sun, 24 Jun 2012 16:50:42 GMT

     [ https://issues.apache.org/jira/browse/CAMEL-5390?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Raul Kripalani updated CAMEL-5390:
----------------------------------

    Description: 
Imagine a sequential JMS invocation across Camel routes, such as the following:

{code}
<route>
    <from uri="timer:foo?fixedRate=true&amp;period=10000" />
    <inOut uri="activemq:queue:test1" />
    <inOut uri="activemq:queue:test2" />
</route>
{code}

The camel-jms consumer listening on test1 will set the {{JMSCorrelationID}} header to the
{{JMSMessageID}} for that exchange.

When the response returns to the route and sent to test2, this consumer also uses the *same*
{{JMSCorrelationID}} as before.

IMHO, this behaviour is incorrect from the pragmatic perspective. Why should test2 reuse the
Message ID from test1 after all?

Despite that, the correlation works for simple cases, but for complex routing with parallel
JMS exchanges, it doesn't because the correlation ID is not unique for each exchange. Consider
the following splitter:
\\
{code}
        <route>
            <from uri="timer:foo?fixedRate=true&amp;period=10000" />
            <setBody><constant>1,2,3,4,5</constant></setBody>
            <inOut uri="activemq:queue:test1" />
            <split parallelProcessing="true">
                <tokenize token="," />
                <inOut uri="activemq:queue:test2" />
            </split> 
            <to uri="log:Finished?showAll=true" />
        </route>
        
        <route>
            <from uri="activemq:queue:test1" />
            <to uri="log:Received?showAll=true" />
        </route>
 
         <route>
            <from uri="activemq:queue:test2" />
            <to uri="log:Received?showAll=true" />
            <setBody><constant>reply</constant></setBody>
            <delay><constant>100</constant></delay>
        </route>
{code}

There are several solutions here:

# Remove the JMSCorrelationID header before each request to test2
# Enable the useMessageIDAsCorrelationID option on both producer and consumer endpoints
# Create a new option to assign a unique ID to the JMSCorrelationID header

The downside of 2 is that the message ID is only assigned after the JMS dispatch, so the component
updates the correlation map with the final JMSMessageID after the JMS dispatch. Also, camel-jms
is also prepared to handle cases where the reply comes in before the map update has occurred,
by waiting 5 seconds if an unknown correlation ID is received, blocking the receipt of any
further replies. You will agree with me this is inefficient and flaky for enterprise deployments.

That's where the upside of 3 lies. By expressly setting a static unique ID as the correlation
ID, we get rid of these race conditions and inefficiencies.

I propose calling this option 'assignUniqueCorrelationID'.

I'm happy to submit a patch for this, targeting 2.10.1.


  was:
Imagine a sequential JMS invocation across Camel routes, such as the following:

{code}
<route>
    <from uri="timer:foo?fixedRate=true&amp;period=10000" />
    <inOut uri="activemq:queue:test1" />
    <inOut uri="activemq:queue:test2" />
</route>
{code}

The camel-jms consumer listening on test1 will set the {{JMSCorrelationID}} header to the
{{JMSMessageID}} for that exchange.

When the response returns to the route and sent to test2, this consumer also uses the *same*
{{JMSCorrelationID}} as before.

IMHO, this behaviour is incorrect from the pragmatic perspective. Why should test2 reuse the
Message ID from test1 after all?

Despite that, the correlation works for simple cases, but for complex routing with parallel
JMS exchanges, it doesn't because the correlation ID is not unique for each exchange. Consider
the following splitter:
\\
{code}
        <route>
            <from uri="timer:foo?fixedRate=true&amp;period=10000" />
            <setBody><constant>1,2,3,4,5</constant></setBody>
            <inOut uri="activemq:queue:test1" />
            <split parallelProcessing="true">
                <tokenize token="," />
                <inOut uri="activemq:queue:test2" />
            </split> 
            <to uri="log:Finished?showAll=true" />
        </route>
        
        <route>
            <from uri="activemq:queue:test1" />
            <to uri="log:Received?showAll=true" />
        </route>
 
         <route>
            <from uri="activemq:queue:test2" />
            <to uri="log:Received?showAll=true" />
            <setBody><constant>reply</constant></setBody>
            <delay><constant>100</constant></delay>
        </route>
{code}

There are several solutions here:

# Remove the JMSCorrelationID header before each request to test2
# Enable the useMessageIDAsCorrelationID option to true on both producer and consumer
# Create a new option to assign a unique ID to the JMSCorrelationID header

The downside of 2 is that the message ID is only assigned after the JMS dispatch, so the component
updates the correlation map with the final JMSMessageID after the JMS dispatch. Also, camel-jms
is also prepared to handle cases where the reply comes in before the map update has occurred,
by waiting 5 seconds if an unknown correlation ID is received, blocking the receipt of any
further replies. You will agree with me this is inefficient and flaky for enterprise deployments.

That's where the upside of 3 lies. By expressly setting a static unique ID as the correlation
ID, we get rid of these race conditions and inefficiencies.

I propose calling this option 'assignUniqueCorrelationID'.

I'm happy to submit a patch for this, targeting 2.10.1.


    
> Option to assign unique correlation ID to JMS messages
> ------------------------------------------------------
>
>                 Key: CAMEL-5390
>                 URL: https://issues.apache.org/jira/browse/CAMEL-5390
>             Project: Camel
>          Issue Type: Improvement
>          Components: camel-activemq
>    Affects Versions: 2.10.0
>            Reporter: Raul Kripalani
>            Assignee: Raul Kripalani
>             Fix For: 2.10.1
>
>
> Imagine a sequential JMS invocation across Camel routes, such as the following:
> {code}
> <route>
>     <from uri="timer:foo?fixedRate=true&amp;period=10000" />
>     <inOut uri="activemq:queue:test1" />
>     <inOut uri="activemq:queue:test2" />
> </route>
> {code}
> The camel-jms consumer listening on test1 will set the {{JMSCorrelationID}} header to
the {{JMSMessageID}} for that exchange.
> When the response returns to the route and sent to test2, this consumer also uses the
*same* {{JMSCorrelationID}} as before.
> IMHO, this behaviour is incorrect from the pragmatic perspective. Why should test2 reuse
the Message ID from test1 after all?
> Despite that, the correlation works for simple cases, but for complex routing with parallel
JMS exchanges, it doesn't because the correlation ID is not unique for each exchange. Consider
the following splitter:
> \\
> {code}
>         <route>
>             <from uri="timer:foo?fixedRate=true&amp;period=10000" />
>             <setBody><constant>1,2,3,4,5</constant></setBody>
>             <inOut uri="activemq:queue:test1" />
>             <split parallelProcessing="true">
>                 <tokenize token="," />
>                 <inOut uri="activemq:queue:test2" />
>             </split> 
>             <to uri="log:Finished?showAll=true" />
>         </route>
>         
>         <route>
>             <from uri="activemq:queue:test1" />
>             <to uri="log:Received?showAll=true" />
>         </route>
>  
>          <route>
>             <from uri="activemq:queue:test2" />
>             <to uri="log:Received?showAll=true" />
>             <setBody><constant>reply</constant></setBody>
>             <delay><constant>100</constant></delay>
>         </route>
> {code}
> There are several solutions here:
> # Remove the JMSCorrelationID header before each request to test2
> # Enable the useMessageIDAsCorrelationID option on both producer and consumer endpoints
> # Create a new option to assign a unique ID to the JMSCorrelationID header
> The downside of 2 is that the message ID is only assigned after the JMS dispatch, so
the component updates the correlation map with the final JMSMessageID after the JMS dispatch.
Also, camel-jms is also prepared to handle cases where the reply comes in before the map update
has occurred, by waiting 5 seconds if an unknown correlation ID is received, blocking the
receipt of any further replies. You will agree with me this is inefficient and flaky for enterprise
deployments.
> That's where the upside of 3 lies. By expressly setting a static unique ID as the correlation
ID, we get rid of these race conditions and inefficiencies.
> I propose calling this option 'assignUniqueCorrelationID'.
> I'm happy to submit a patch for this, targeting 2.10.1.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message