activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Raul Kripalani <r...@evosent.com>
Subject Re: Broker latency spike
Date Tue, 13 Nov 2012 21:56:46 GMT
Are messages persistent? If so, the latency could be easily introduced by
the rotation of the message journal, which triggers a Disk IO operation to
pre-allocate a new 32mb file before it can ACK the message to the producer.
Check out the options here [1].

[1] http://activemq.apache.org/kahadb.html

Regards,
Raúl K.

On Tue, Nov 13, 2012 at 3:57 PM, zackhasit <zackhasit@gmail.com> wrote:

>
> Hi, I ran the producer/consumer which is delivered as part of the example.
> I
> added code to show difference between JMSActiveMQBrokerInTime and
> JMSActiveMQBrokerOutTime. It remains zero for 90% of the time but goes
> higher every now and then. I suspected garbage collection and now the
> broker
> parameters are :
>
> /tools/java/jdk1.6.0_29/bin/java -Xms1G -Xmx1G -XX:+UseParNewGC
> -XX:+UseConcMarkSweepGC -XX:+CMSParallelRemarkEnabled -XX:SurvivorRatio=20
> -XX:ParallelGCThreads=8 -Dorg.apache.activemq.UseDedicatedTaskRunner=true
> -Djava.util.logging.config.file=logging.properties
> -Dcom.sun.management.jmxremote
> -Dactivemq.classpath=/apache-activemq-5.5.1/conf;
> -Dactivemq.home=/apache-activemq-5.5.1
> -Dactivemq.base=/apache-activemq-5.5.1 -jar
> /apache-activemq-5.5.1/bin/run.jar start
>
> But I dont see any improvement. I looked at other options like async send
> etc but that did not help. Consumers,broker and producers are all on same
> server and there is enough CPU and memory (no h/w bottlenecks as far as I
> can see...) and examples are using CMS to connect to broker. Any
> recommendation on what to try ? Since this is a standard example I am
> hoping
> folks here are already aware of this ?
>
>
>
> --
> View this message in context:
> http://activemq.2283324.n4.nabble.com/Broker-latency-spike-tp4659255.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message