You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by metaldork <me...@email.com> on 2008/02/27 00:20:12 UTC
need help interpreting jmx values
hi. i'm preparing to migrate to jms for oracle toplink cache synching and am
currently doing some configuration and load testing. my test is:
4 brokers in distributed topic configuration using dynamic discovery
a junit test that starts up 10 threads, each connecting as a consumer using
dynamic discovery
a junit test that starts up 10 threads, each connecting as a producer
when i first start up the brokers but no consumers or producers, i see the
consumer count in each is 1. assuming this is each broker connecting to
another as a client, i can accept this.
what doesn't make sense is when i start my consumers up, i get a total of 20
consumers spread across the brokers when i would expect 12.
secondly, the #s displayed as enqueue and dispatch counts don't add up. in
my last test run, 3 of the servers recorded 6155 enqueues but one only 1822.
furthermore, when i divide the dispatch count by the enqueue count, it
sometimes, but not always, adds up to the # of consumers.
finally, in my test i record the time i send a message to a topic and when
each consumer pulls it out i calculate the total time it spent in the
broker. usually it's a matter of milliseconds. but ocassionally, i'll get
a pause, then a large load will be pulled out that had been in the broker
for many seconds or even minutes. then another pause, then another load
that had been in the broker even longer. all my consumer does is pull out a
message and display the time so i don't think it's that the consumers can't
keep up.
anyhoo, i'm perplexed and i can't go live with this until everybody is
confident so if anybody can help me out i'd greatly appreciate it.
--
View this message in context: http://www.nabble.com/need-help-interpreting-jmx-values-tp15702471s2354p15702471.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
Re: need help interpreting jmx values
Posted by metaldork <me...@email.com>.
ok so i think i've figured out the consumer count thing. but first, what was
confusing me was that my ops guy had brought up a test instance of the
application this is intended for on the same network without letting me
know. so once that was taken down it became more clear. here's what i
gather:
i have 4 brokers configured for max 1 network hop. so each broker has to be
a consumer to every other broker that has a consumer listening on that
topic. my initial thought was that each broker just became a consumer to
any one of the other brokers.
each broker only becomes a consumer to a broker that has a non-broker
consumer. so if all 4 brokers get a non-broker consumer, each has 3 broker
consumers (12 across the network) plus 10 non-broker consumers for a total
of 22.
now, having said that, is it more efficient to allow a max of 3 network
hops, resulting in fewer consumers but more potential distance for a message
to travel?
also, getting rid of this bogus consumer seems to have fixed the problem
with pauses in message dequeueing
James.Strachan wrote:
>
> Networks work by brokers acting as consumers on other brokers which
> probably explains the unexpected consumer counts?
>
> On 26/02/2008, metaldork <me...@email.com> wrote:
>>
>> hi. i'm preparing to migrate to jms for oracle toplink cache synching
>> and am
>> currently doing some configuration and load testing. my test is:
>>
>> 4 brokers in distributed topic configuration using dynamic discovery
>> a junit test that starts up 10 threads, each connecting as a consumer
>> using
>> dynamic discovery
>> a junit test that starts up 10 threads, each connecting as a producer
>>
>> when i first start up the brokers but no consumers or producers, i see
>> the
>> consumer count in each is 1. assuming this is each broker connecting to
>> another as a client, i can accept this.
>>
>> what doesn't make sense is when i start my consumers up, i get a total
>> of 20
>> consumers spread across the brokers when i would expect 12.
>>
>> secondly, the #s displayed as enqueue and dispatch counts don't add up.
>> in
>> my last test run, 3 of the servers recorded 6155 enqueues but one only
>> 1822.
>> furthermore, when i divide the dispatch count by the enqueue count, it
>> sometimes, but not always, adds up to the # of consumers.
>>
>> finally, in my test i record the time i send a message to a topic and
>> when
>> each consumer pulls it out i calculate the total time it spent in the
>> broker. usually it's a matter of milliseconds. but ocassionally, i'll
>> get
>> a pause, then a large load will be pulled out that had been in the
>> broker
>> for many seconds or even minutes. then another pause, then another load
>> that had been in the broker even longer. all my consumer does is pull
>> out a
>> message and display the time so i don't think it's that the consumers
>> can't
>> keep up.
>>
>> anyhoo, i'm perplexed and i can't go live with this until everybody is
>> confident so if anybody can help me out i'd greatly appreciate it.
>>
>> --
>> View this message in context:
>> http://www.nabble.com/need-help-interpreting-jmx-values-tp15702471s2354p15702471.html
>> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>>
>>
>
>
> --
> James
> -------
> http://macstrac.blogspot.com/
>
> Open Source Integration
> http://open.iona.com
>
>
--
View this message in context: http://www.nabble.com/need-help-interpreting-jmx-values-tp15702471s2354p15724682.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
Re: need help interpreting jmx values
Posted by James Strachan <ja...@gmail.com>.
Networks work by brokers acting as consumers on other brokers which
probably explains the unexpected consumer counts?
On 26/02/2008, metaldork <me...@email.com> wrote:
>
> hi. i'm preparing to migrate to jms for oracle toplink cache synching and am
> currently doing some configuration and load testing. my test is:
>
> 4 brokers in distributed topic configuration using dynamic discovery
> a junit test that starts up 10 threads, each connecting as a consumer using
> dynamic discovery
> a junit test that starts up 10 threads, each connecting as a producer
>
> when i first start up the brokers but no consumers or producers, i see the
> consumer count in each is 1. assuming this is each broker connecting to
> another as a client, i can accept this.
>
> what doesn't make sense is when i start my consumers up, i get a total of 20
> consumers spread across the brokers when i would expect 12.
>
> secondly, the #s displayed as enqueue and dispatch counts don't add up. in
> my last test run, 3 of the servers recorded 6155 enqueues but one only 1822.
> furthermore, when i divide the dispatch count by the enqueue count, it
> sometimes, but not always, adds up to the # of consumers.
>
> finally, in my test i record the time i send a message to a topic and when
> each consumer pulls it out i calculate the total time it spent in the
> broker. usually it's a matter of milliseconds. but ocassionally, i'll get
> a pause, then a large load will be pulled out that had been in the broker
> for many seconds or even minutes. then another pause, then another load
> that had been in the broker even longer. all my consumer does is pull out a
> message and display the time so i don't think it's that the consumers can't
> keep up.
>
> anyhoo, i'm perplexed and i can't go live with this until everybody is
> confident so if anybody can help me out i'd greatly appreciate it.
>
> --
> View this message in context: http://www.nabble.com/need-help-interpreting-jmx-values-tp15702471s2354p15702471.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>
>
--
James
-------
http://macstrac.blogspot.com/
Open Source Integration
http://open.iona.com