activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark Gellings (JIRA)" <jira+amq...@apache.org>
Subject [jira] Reopened: (AMQNET-159) Make failover protocol optionally asynchronously connect to sockets
Date Thu, 22 Oct 2009 20:25:53 GMT

     [ https://issues.apache.org/activemq/browse/AMQNET-159?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Mark Gellings reopened AMQNET-159:
----------------------------------


I reviewed the changes and found one difference.

In FailoverTrasport.doConnect()

if(_asyncConnect)
{
	// now wait for transport to be populated, but timeout eventually
	allDone.WaitOne(_asyncTimeout, false);
}

you are passing false into an overloaded WaitOne method.  I was not using this method.  I
wasn't passing in false just _asyncTimeout.  Is there a specific reason this was done?  If
so, how come?  If not, what are the implications of this?

Also, would it make sense to make the async connection the default behavior?

> Make failover protocol optionally asynchronously connect to sockets
> -------------------------------------------------------------------
>
>                 Key: AMQNET-159
>                 URL: https://issues.apache.org/activemq/browse/AMQNET-159
>             Project: ActiveMQ .Net
>          Issue Type: New Feature
>          Components: ActiveMQ
>    Affects Versions: 1.1.0
>         Environment: XP, Windows Server 2008, Sql Server 2008
>            Reporter: Mark Gellings
>            Assignee: Jim Gomes
>             Fix For: 1.2.0
>
>         Attachments: FailoverAsyncConnect.patch
>
>
> The current failover transport synchronously attempts to socket connect to a broker.
 Preferably we would first socket connect trying the master address in the failover address
to avoid a socket connection timeout.  When the first tcp address in the failover address
is the slave, its connectors are not started, and .NET takes around 20 - 25s to timeout trying
to socket connect to the slave.  Thus the client is waiting 20-25s to submit a message when
the session is first established.
> Since the first tcp address in the failover address is never guaranteed to be the master,
I've coded an enhancement that optionally allows to asynchronously socket connect utilizing
all the tcp addresses in the failover address.  I have a patch you can apply.
> The following failover address is an example of how to use the async connect.
> private static Uri _uri = new Uri("failover:(tcp://WAMQDEV1.qg.com:10083,tcp://SXMQDEV1.qg.com:10083)?AsyncConnect=true&AsyncTimeout=20000");
> This will cause the failover transport to attempt connecting to WAMQDEV1 and SXMQDEV1
at the same time to avoid the impact of any socket connect timeouts.  First broker connected
to becomes the transport utilized.  If WAMQDEV1 broker is not running or is running the slave
broker then if SXMQDEV1 is running the master broker the client will connect to SXMQDEV1.
> AsyncTimeout is the max time in milliseconds to wait for a transport to be initialized.
 So if both brokers are down, in the example above, after 20000 milliseconds the normal retry
logic will kick in.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message