Mailing-List: contact issues-help@camel.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@camel.apache.org
Date: Mon, 3 Nov 2014 21:35:35 +0000 (UTC)
From: "Aaron Whiteside (JIRA)" <jira@apache.org>
To: issues@camel.apache.org
Message-ID: <JIRA.12752533.1415050325000.401983.1415050535857@Atlassian.JIRA>
In-Reply-To: <JIRA.12752533.1415050325000@Atlassian.JIRA>
References: <JIRA.12752533.1415050325000@Atlassian.JIRA>
 <JIRA.12752533.1415050325424@arcas>
Subject: [jira] [Updated] (CAMEL-7994) SJMSComponent effectively ignores
 setConnectionCount()
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/CAMEL-7994?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Aaron Whiteside updated CAMEL-7994:
-----------------------------------
    Description: 
SJMSComponent effectively ignores setConnectionCount()

This is because the JmsProducer, by default, pre-populates sessions+producers on Producer.doStart(), which is called by the CamelContext in a single thread and the way the ConnectionFactoryResource is configured, it returns the most recently idle connection (So really it always returns the first connection opened).

This patch attached makes the ConnectionFactoryResource effectively return connections in a round-robbin manor. Unit test is included.

This is important because TCP has a certain overhead per socket which makes it impossible to achieve much more than 6k~ messages a second, regardless of the CPU and IO resources available to a broker. To fully utilize a broker multiple connections must be opened.

This is also important if your ConnectionFactory represent a cluster (Like in the case of HornetQ), where subsequent connections from the factory are balanced across nodes in the cluster.

  was:
SJMSComponent effectively ignores setConnectionCount()

This is because the JmsProducer, by default, pre-populates sessions+producers on Producer.doStart(), which is called by the CamelContext in a single thread and the way the ConnectionFactoryResource is configured, it returns the most recently idle connection (So really it always returns the first connection opened).

This patch attached makes the ConnectionFactoryResource effectively return connections in a round-robbin manor. Unit test is included.

This is important because TCP has a certain overhead per socket which makes it impossible to achieve much more than 6k~ messages a second, regardless of the CPU and IO resources available to a broker. To fully utilize a broker multiple connections must be opened.

This is also important if your ConnectionFactory represent a cluster (Like in the case of HornetQ), where subsequent connections are balanced across nodes in the cluster.


> SJMSComponent effectively ignores setConnectionCount()
> ------------------------------------------------------
>
>                 Key: CAMEL-7994
>                 URL: https://issues.apache.org/jira/browse/CAMEL-7994
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-sjms
>    Affects Versions: 2.14.0
>            Reporter: Aaron Whiteside
>         Attachments: unnamed4.patch
>
>
> SJMSComponent effectively ignores setConnectionCount()
> This is because the JmsProducer, by default, pre-populates sessions+producers on Producer.doStart(), which is called by the CamelContext in a single thread and the way the ConnectionFactoryResource is configured, it returns the most recently idle connection (So really it always returns the first connection opened).
> This patch attached makes the ConnectionFactoryResource effectively return connections in a round-robbin manor. Unit test is included.
> This is important because TCP has a certain overhead per socket which makes it impossible to achieve much more than 6k~ messages a second, regardless of the CPU and IO resources available to a broker. To fully utilize a broker multiple connections must be opened.
> This is also important if your ConnectionFactory represent a cluster (Like in the case of HornetQ), where subsequent connections from the factory are balanced across nodes in the cluster.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)