activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bruce Snyder <bruce.sny...@gmail.com>
Subject Re: Persistence Behavior
Date Wed, 16 Sep 2009 15:22:31 GMT
On Tue, Sep 15, 2009 at 4:48 PM, dbczak <zak@dbconcert.com> wrote:
>
> Yes I'm using failover.  The problem is that if I don't set the
> maxReconnectAttempts, it waits forever when I call send and the connection
> is down.  But if I set the maxReconnectAttempts, then after it reaches that
> max number of attempts, it never connnects again.  Shouldn't I be able to
> attempt a send, and continue processing if send fails, and then when the
> consumer comes back up, the reconnect should occur and the missed messages
> should be processed.

First, let's deal with the connection issue. If the producer is using
the failover protocol and the connection fails the failover transport
will attempt to automatically reconnect to the broker. There are a few
properties to fine-tune the reconnection listed here:

http://activemq.apache.org/failover-transport-reference.html

I just tested this using the examples that come with ActiveMQ in the
example dir. Here are the steps I used:

1) In terminal one, start ActiveMQ using the following command:

$ ./bin/activemq

2) In terminal two, cd into the example dir and running the following command:

$ ant producer -D"url=failover:(tcp://localhost:61616)"

The producer should begin sending messages. By default, it will send
2000 messages.

3) Go to terminal one and kill the broker using ctrl-c
4) Go to terminal two and notice the following error in the producer:

[java] 08:58:55 WARN  Transport failed to tcp://localhost:61616 ,
attempting to automatically reconnect due to: java.io.EOFException

5) Go to terminal one and start up ActiveMQ again.
6) Go to terminal two and the producer will output the following
message before it begins sending messages again:

[java] 08:59:06 INFO  Successfully reconnected to tcp://localhost:61616

This is how the failover transport should operate. If you are not
seeing the same behavior, then this needs to be addressed to determine
why.

Second, the MessageProducer.send() method is a synchronous (i.e.,
blocking) call which requires an active connection to a broker. There
is no message buffering capability built in to the message producer to
cushion against a broker failure. This is typically where folks use
the failover transport. One solution is to use an embedded broker in
your application and let it forward messages to the remote broker via
network of brokers:

http://activemq.apache.org/networks-of-brokers.html

By embedding a broker in the Java application and networking it with a
remote broker, the ActiveMQ network transport will be used to forward
messages between the two brokers. Using this scenario, your
application will send locally to the embedded broker using the vm
transport (http://activemq.apache.org/vm-transport-reference.html)
which is very fast. The messages will be held in that broker until
there is demand on the remote broker for those messages (i.e., an
active consumer on the same destination on the remote broker). When
demand is created for those messages on the remote broker, the local
broker will forward them along to be consumed.

Bruce
-- 
perl -e 'print unpack("u30","D0G)U8V4\@4VYY9&5R\"F)R=6-E+G-N>61E<D\!G;6%I;\"YC;VT*"
);'

ActiveMQ in Action: http://bit.ly/2je6cQ
Blog: http://bruceblog.org/
Twitter: http://twitter.com/brucesnyder

Mime
View raw message