A few qns to help with the situation:
1. what's the hardware where you're running the broker - #machines, type of
machines, config, etc?
2. how much heap are you giving the amq processes? anything else running on the
broker nodes or are they dedicated to amq?
3. are messages persistent? if so,
You could also just turn on gc logging on the AMQ nodes and use the log file
for analysis on the same nodes or rsync it out to another node and simulate a
gc probe/agent.
On Jun 25, 2014, at 10:12, Rodrigo Ramos wrote:
Heap dump its for that. A good tool for analize heap dums and get statics
i
Without knowing much about the domain, now that doesn't seem like the
best way to attack this problem. If you were hoping to do any sort of
secondly or few-secondly moving windows, the rpc/remoteness alone
would negate the use of such a system. A remote system for rate
throttling is better suited f
http://sensatic.net
ActiveMQ in Action: http://www.manning.com/snyder/
On Thu, Jun 27, 2013 at 5:38 PM, Gaurav Sharma
wrote:
> Thanks Dejan. So should I wait for a v5.8.1 being cut in the near future?
>
> On Jun 27, 2013, at 8:34, Dejan Bosanac wrote:
>
> Hi Gaurav,
>
> the
...@redhat.com
Twitter: @dejanb
Blog: http://sensatic.net
ActiveMQ in Action: http://www.manning.com/snyder/
On Thu, Jun 27, 2013 at 5:30 PM, Gaurav Sharma
wrote:
> I am planning to upgrade our v5.6 clusters to 5.9. Please advise what's
> the tentative release date for 5.9? If it
I am planning to upgrade our v5.6 clusters to 5.9. Please advise what's the
tentative release date for 5.9? If it's close, I can point all non-prod
clusters to 5.9 snapshot and test/wait for the release to be cut before
updating prod.
Thank you,
-gaurav
I'd be interested in discussing this, as well though we're not there yet but
going to be in the next month or so. You do all have the basic args, yes -
start/max heap (same values), print gc logs, tstamps, dump on oom, compressed
oops might be on by default, etc?
Sent from my iPhone
On Apr 2,
So, here're a few of my observations and suggestions based on the provided info:
1. use a newer java than v1.7.0_03
2. switch to amq v5.7 or v5.8
3. keep xmx and xms the same and preferably a multiple of 512 megs
4. where are you getting server grade machines with 2 and 15 gig ram
5. when you notic
Believe we'll have to backport this to v5.8.1 as well. I'm holding out for
v5.8.1 to upgrade prod clusters from v5.6.0. Thank you.
On Mar 27, 2013, at 8:52, Timothy Bish wrote:
On 03/27/2013 11:45 AM, iamactivemquser wrote:
> Hi all,
>
> We want to upgrade from ActiveMQ 5.5.1 to 5.8.0. We are
Wish I could put up some of Hiram's benchmark ActiveMQ tests' perf stats up
here from even one of our dev clusters (forget about stress or prod env's) but
company policy doesn't permit I share. One of my prod clusters does apple push
notifications among other things - it has been up and invisibl
up around the AMQ site on PFC, system limits, slow
> > >subscribers and cursors to get a better insight. Adjusting these is
> > >straightforward and yield quick wins.
> > >
> > >Also, post a thread dump the next time, as someone else already said.
> &g
a clue what you are talking about.
>
>
> From: Gaurav Sharma [via ActiveMQ]
> [mailto:ml-node+s2283324n4662992...@n4.nabble.com]
> Sent: Thursday, February 07, 2013 10:57 AM
> To: Masanz, Don
> Subject: Re: Stuck Messages
>
> Forgot to mention - since you know what qu
the broker so I can
auto-markdown bad apple broker nodes on the client side.
On Feb 7, 2013, at 8:40, Gaurav Sharma wrote:
> Folks - when you notice the broker in a 'stuck' state, please take a thread
> and heap dump before killing the broker process and share with devs on he
Folks - when you notice the broker in a 'stuck' state, please take a thread and
heap dump before killing the broker process and share with devs on here.
Also:
1. What broker topology are you using?
2. How many producers/consumers?
3. Are you pooling connections, producers, sessions?
4. How l
Okay, so, you have been able to somewhat correlate the slow connection
establishment to load?
Are you using tcp or nio transport? And for the host running the broker, how
much is the:
a) RAM
b) Heap for the JVM
c) SystemUsage settings
I suspect your broker jvm is undergoing a full gc.
On Jan
It could be a number of things. But first, a couple questions:
1. how does your network look like - 2/3 level switches in the DC, flat or
something else? Where are the producers deployed relative to the broker
clusters? Do you have tcp keep-alive turned on, spl routing, firewalls, etc?
2. any
Yes, see http://activemq.apache.org/rest.html
On Nov 20, 2012, at 3:03, James Green wrote:
> Trying to whip up a quick script to test performance.
>
> It would be really useful to call a URI to purge/delete the queue I'm using
> between iterations without needing to manually drain it.
>
> Is
Mind sharing the messaging topology and how the broker is getting used by the
app?
On Nov 6, 2012, at 8:49, Aravind Chennuru wrote:
> Hi All,
>
> I am facing a problem on some of our production servers with ActiveMQ locking
> the threads and the messages are no more sent to the consumers. Did
Some points/observations:
1. Do you have NOB config for HA reasons only? Because if you are
immediately consuming published messages, what's the benefit of paying
the price of broker-to-broker message transfer.. there are other means
to achieve availability and message durability without having to
g.properties -Xss256k
> ---
> There are 1 CPU of 4 kernels and 6G memory on this machine, and no
> other app is running.
>
>
> 2012/9/21 Gaurav Sharma [via ActiveMQ]
> :
>> How much h
How much heap have we given to the broker and what else is running on the same
machine? Also, just fire up jconsole, attach to the process and watch
allocations, threads and cpu - should give a clue. v5.6 is the current release,
so, upgrade if possible.
On Sep 20, 2012, at 23:34, jockeyyan wro
So, are A, B, C brokers on three separate JVM's and looks like (from your
description) that their producers are local on the same machine? By
"instances", you do mean brokers, right?
With this 3 broker config, what use case are you trying to solve by routing
message from A to B or C? Like you s
You can use the message group property or just roll your own k-v pair property
and a selector on the consumer side.
Btw, what do you mean when you say "AMQ A, B, C"? The answer could be (in basic
terms) any combination of broker, destination, producer or consumer.
On Sep 20, 2012, at 5:00, mad
Thanks Christian. I will give it a spin on one of our clusters in the coming
week.
On Sep 4, 2012, at 14:41, ceposta wrote:
> I've recently seen questions about benchmarking ActiveMQ.
>
> I've added a new blog post that describes one way to do so:
>
> Benchmarking your JMS layer with an open
imveMQ via JMX?
>
>
>
>
>
> rmn190
>
> From: Gaurav Sharma
> Date: 2012-09-09 00:54
> To: users@activemq.apache.org
> Subject: Re: configure ActiveMQ with Zookeeper?
> It is doable but zk's real power is in dynamic config. mgmt. So, if you
> review the
It is doable but zk's real power is in dynamic config. mgmt. So, if you review
the xml schema, a lot of the elements of activemq.xml are not dynamically
reconfigurable and require a restart. So, why introduce this other component
for configuration anyway? Also, AMQ already exposes rich JMX based
I think it would make a
>> sensible enhancement.
>>
>> [1]
>> http://svn.apache.org/viewvc/activemq/trunk/activemq-web/src/main/java/org/apache/activemq/web/RemoteJMXBrokerFacade.java?view=markup
>>
>> On 5 September 2012 06:18, Gaurav Sharma wrote:
>>&
For the JDBC Master Slave config, just like there exists the Failover
Transport for the broker url, what is the best way to specify the JMX
hosts in a properties file:
jmxUrl=service:jmx:rmi:///jndi/rmi://:1616/jmxrmi, \
service:jmx:rmi:///jndi/rmi://:1616/jmxrmi, \
service:jm
Hmm, actually, if you do setMessageListener(null) on
the DefaultMessageListenerContainer, it will throw an
IllegalArgumentException in checkMessageListener(), so, that shouldn't work.
The stop(), start(), isRunning() methods do seem like a good option to
pause the listener via spring. The other wa
ug 27, 2012 at 12:38 AM, Tlholoe, Peter wrote:
> Hi Gaurav,
> I did take thread dumps during hang, and after the fact. I have attached
> 10 thread dump file. T1dump -t7dump are during the hang and t7dump to
> t10dump are after the hang period.
>
> -Original Message
When the process hangs, please take a thread dump (kill -3) and pastebin/gist
and share so folks on here can help. Also, how did you determine that it hangs
- no log activity, jmx response, etc?
By the way, 1-2gigs of heap is not very large these days not that you should
just up it but just fyi
hing with load-balance, right?
>
>
>
> On Fri, Aug 24, 2012 at 4:37 PM, Gaurav Sharma [via ActiveMQ] <
> ml-node+s2283324n4655527...@n4.nabble.com> wrote:
>
> > Another way to look at it is in terms of the sticky-session analogy
> though
> > don't be confused
Another way to look at it is in terms of the sticky-session analogy though
don't be confused by the concept of the 'session'. The keyword of interest
is 'sticky'.
On Fri, Aug 24, 2012 at 1:31 AM, rmn190 wrote:
> in http://activemq.apache.org/message-groups.html, there is a saying:
> Message
> Gr
This is a difficult question to answer but briefly, it will depend on the
services' design, state management of the system, concurrency-model, the
domain and what the services will be doing. I presume they will have
different behaviors with a mixture of cpu-bound, diskIO-bound,
networkIO-bound, etc
So, instead of having N brokers (1 per consumer), why not have N
destinations (1 per consumer)? If possible, further simplify it to less
than N destinations if you can shard/group your messages (don't know your
domain) and just use message properties and selectors to filter them at the
time of retr
So, that request-response recipe uses temporary destinations to route
messages. Your use-case is that of request-throttling and load-balancing
the 'service'. Why not use something designed for that purpose like nginx
and its http_limit_conn module? ActiveMQ does have the producer
flow-control featu
To register an HttpCallback, you can create a class with a simple method
like below and just register the messagelistener at message publish time
(since your producers seem to know the message route upfront). This can be
called in a pooled thread on the client side:
void invokeEndpoint() {
So you can always register an http-callback in the garb of a message listener
during the time of msg publication. When the client execs your listener and
runs the callback, it can do http post/get, etc to whatever other broker url
you want to route the msg to.
I am not sure what multiple broke
Per your original message: "Brokers at both producer and consumer are
running as java embedded brokers"
So, you don't need two embedded brokers for passing the same message
between a producer and a consumer. Just run a single broker on a machine
that is accessible to the unix-producer and the wind
Hi Poorva, you can do with just a single broker (on the same jvm, forked or
remote) and producers and consumers as 'clients' of the broker.
For absolute simplicity, consider the broker as a pipe/container with
producer-push and consumer-pull or consumer-push (via messagelistener) like:
producer ->
In the ACTIVEMQ_LOCK table, is the null value in BROKER_NAME to be
expected? I read some old threads about this but couldn't find a definitive
answer from one of the committers. While it is a minor inconvenience for
us, at runtime, without scanning logs of the broker machines, there is a no
way to
twitter: @dejanb
> blog: http://www.nighttale.net
> ActiveMQ in Action: http://www.manning.com/snyder/
>
>
> On Fri, Jul 27, 2012 at 12:26 PM, Gaurav Sharma
> wrote:
> > Dejan - thanks for the detailed explanation. Based on this, I went back
> and
> > figured
will reconnect
> to host2.
>
>
> Regards
> --
> Dejan Bosanac
> Senior Software Engineer | FuseSource Corp.
> dej...@fusesource.com | fusesource.com
> skype: dejan.bosanac | twitter: @dejanb
> blog: http://www.nighttale.net
> ActiveMQ in Action: http://www.manning.com/sn
It seems like with a failover transport configuration
(failover:(nio:host1:port1,nio:host2:port2)?randomize=false) and a
PooledConnectionFactory, the client-side still tries to create a new
connection per thread instead of fetching from the pool of connections.
With 'randomize' flag turned off, I w
This seems like an interesting problem and assuming it is no longer ntp
related, it seems to hint at a livelock but cannot say anything
definitively without setup information. Can you pastebin or gist jstack -F
output and also your config's? This of interest to me and others
planning to switch to
I think you forgot to do: connection.start();
Also, I have the pool size configured on the pooledconnectionfactory via
setMaxConnections() and setMaximumActive() for tighter control.
On Mon, Jul 23, 2012 at 9:43 AM, cwash wrote:
> BTW, using version 5.6
>
>
>
> --
> View this message in context:
eremy Levy wrote:
> Thanks again for the quick response. That's good info. What volume of
> messages are you processing? I'm doing about 25 million a day.
>
>
> Jeremy
>
> On Fri, Jul 20, 2012 at 2:54 PM, Gaurav Sharma
> wrote:
>
>> Connection factory
on factory around statically? I've played with
> async sending with mixed results.
>
> Jeremy
>
> On Thu, Jul 19, 2012 at 7:53 PM, Gaurav Sharma
> wrote:
>
> > You might wanna use the PooledConnectionFactory wrapper with the
> > ActiveMQConnectionFactory. Th
You might wanna use the PooledConnectionFactory wrapper with the
ActiveMQConnectionFactory. That's what I switched to too. If you have a
listener, you will have to leave the session open as well until msg delivery. I
am using mysql as the msg-store and nio connector transport to the broker.
Als
Hi all,
Is anyone using ActiveMQ with Oracle for persistence and if you have done
some performance benchmarking, can you share some numbers please?
Performance test numbers are fine to share too:
http://activemq.apache.org/activemq-performance-module-users-manual.html
Thank you.
-Gaurav
50 matches
Mail list logo
