Glassfish3 cluster queue browser

Hi there,
I have some web app running in a Gv3 cluster. I'd like to be able to query/browse messages for a given queue for each node in the cluster.
I've read about [message pre-sending|http://openmessaging.blogspot.com/2009/03/consumer-flow-control-and-message.html] and my understanding is that if I just use a QueueBrowser I can only see the messages for the node that handles the current request. This is quite confusing, because each reload might result in a different node and different messages.
It also makes it hard to delete messages from a queue, because with load balancing there is a good chance to pick a different node each request and then the referenced message does not exist (on the node handling the new request).
Any ideas/pointers appreciated.

DerManoMann wrote:
Well, that was just following the (correct) suggestion of finding a more appropriate place to ask.
I wasn't aware that cross posting notes are also required for posts on completely different forums. My sincerer apologies should I have offended anyoneIn your defense I do think that people take offense a little to easily on this topic; care should be taken to separate the real crossposters (who spam the same question on several forums at the same time) and people who just use the internet properly.
What you did was correct IMO, but the next time you may want to post a link to your new discussion here so people can follow it.

Similar Messages

JMS Queue browser for SAP XI J2EE Engine

Hi Gurus,
Does anybody has done this before? I need to develop a custom application which looks into the JMS queues on the SAP J2ee engine and raises an alert in case of the message sits on the queue for long time.
Please help in this regard.
Thanks
Kalyan Veerapaneni

There are couple of free source softwares available which work well with IBM MQseries and at least will give you an idea regarding the browse functionality.A google search should help you.

WL6.1 - browsing and receiving from queues with MDBs

Hi all,
          I have hit on a problem that has been brought up previously without,
          as far as I can tell, any conclusive response.
          I am using a queue which is processed by message driven beans. This
          queue contains jobs which may take a long time to execute and hence it
          is important to allow users to manage the queue. Ideally this
          management should work in the same way as, say, a printer queue where
          it is possible to view queued jobs and cancel them if necessary.
          It has been said that MDBs in WLS6.1 use an internal pipeline which
          [quote from previous thread " QueueBrowser not working in WL6"]
          "uses an optimistic algorithm and pushes the message at the bean
          before it is ready"
          and, sure enough, pointing a queue browser, activated from a JSP, at
          the queue always returns a queue length of zero, whatever the number
          of pooled MDBs and whatever the actual number of messages queued. I
          have worked around this by creating a second 'monitoring' queue not
          listened to by any MDBs and which I then have to manually synchronise
          with the 'real' jobs queue. However, removing jobs does not work
          either (I am storing the job message ID and trying to access the jobs
          queue using this as a selector) since the queue is empty when looked
          at from any client which is not the listening MDB. In contrast with
          the browsing, there seems to be no obvious way around this as I need
          to access the jobs queue itself before I can remove any jobs.
          My intended 'queue management' functionality does not seem
          particularly esoteric to me and hence I am puzzled that this appears
          difficult to implement. I have therefore been wondering whether I have
          been missing any obvious point and I would be grateful to hear of
          anybody else's experiences trying to achieve similar functionality.
          It would also be interesting to hear whether moving to WLS 7.0 would
          change/fix this 'pipelining' behaviour of MDBs.
          Many thanks.
          Peter


Hi Peter,
          Since 6.0, when pipelining was introduced,
          the pipeline size is configurable via the connection
          factory (see "MessagesMaximum"). To tune this setting,
          you will need to configure a
          custom connection factory for the MDB rather than use
          the default one, and specify its name in the xml descriptor.
          I'm a little surprised you did not find reference to this
          setting in your search.
          Regardless, you will not be able to stop the current message in the MDB.
          In the next release (beta December) you can force the current tx to
          rollback through the JTA mbeans. Of course, this won't stop
          the onMessage() from continuing its work, but it will
          guarantee that any of its transactional
          work will be ultimately discarded.
          Since the app server has no control of the thread
          once it is passed to the MDB app (it can't just kill
          the thread), you will need to roll your
          own mechanism to make it stop. One possibility is to have your MDB
          "register" that it is starting with a "manager" application, and then
          have the MDB poll this application to see if the system admin
          has requested that the current job be aborted.
          You will likely find it useful to read through the JMS FAQ. You
          might also find it useful to look at the jsp queue browser app
          posted on dev2dev by Adarsh...
          Tom, BEA
          P.S. The release after next release will contain more advanced message
          management features.
          Peter Case wrote:
          > Hi all,
          >
          > I have hit on a problem that has been brought up previously without,
          > as far as I can tell, any conclusive response.
          > I am using a queue which is processed by message driven beans. This
          > queue contains jobs which may take a long time to execute and hence it
          > is important to allow users to manage the queue. Ideally this
          > management should work in the same way as, say, a printer queue where
          > it is possible to view queued jobs and cancel them if necessary.
          > It has been said that MDBs in WLS6.1 use an internal pipeline which
          >
          > [quote from previous thread " QueueBrowser not working in WL6"]
          > "uses an optimistic algorithm and pushes the message at the bean
          > before it is ready"
          >
          >
          > and, sure enough, pointing a queue browser, activated from a JSP, at
          > the queue always returns a queue length of zero, whatever the number
          > of pooled MDBs and whatever the actual number of messages queued. I
          > have worked around this by creating a second 'monitoring' queue not
          > listened to by any MDBs and which I then have to manually synchronise
          > with the 'real' jobs queue. However, removing jobs does not work
          > either (I am storing the job message ID and trying to access the jobs
          > queue using this as a selector) since the queue is empty when looked
          > at from any client which is not the listening MDB. In contrast with
          > the browsing, there seems to be no obvious way around this as I need
          > to access the jobs queue itself before I can remove any jobs.
          > My intended 'queue management' functionality does not seem
          > particularly esoteric to me and hence I am puzzled that this appears
          > difficult to implement. I have therefore been wondering whether I have
          > been missing any obvious point and I would be grateful to hear of
          > anybody else's experiences trying to achieve similar functionality.
          > It would also be interesting to hear whether moving to WLS 7.0 would
          > change/fix this 'pipelining' behaviour of MDBs.
          >
          > Many thanks.
          >
          > Peter

Browsing routed queues?

Hi,
The JMS provider I am using has the ability to "route" queues and topics - forwarding the messages from one server to another. This comes in handy when messages need to cross a WAN so clients don't have to connect over the WAN. Unfortunately, the provider throws an exception when I try to create a queue browser for one of the routed queues. The exception text makes it clear that they do not allow you to browse routed queues. This conflicts with my understanding of the JMS spec, but I wanted to see what other people thought before approaching the vendor. Should I be able to browse a queue if that queue exists? I had thought the InvalidDestinationException would only be thrown only if I tried to connect to a nonexistent queue.
Thanks for any feedback!

The JMS provider I am using has the ability to
o "route" queues and topics - forwarding the messages
from one server to another. This comes in handy when
messages need to cross a WAN so clients don't have to
connect over the WAN. Unfortunately, the provider
throws an exception when I try to create a queue
browser for one of the routed queues. The exception
text makes it clear that they do not allow you to
browse routed queues. This conflicts with my
understanding of the JMS spec, but I wanted to see
what other people thought before approaching the
vendor. Should I be able to browse a queue if that
queue exists? I had thought the
InvalidDestinationException would only be thrown only
if I tried to connect to a nonexistent queue.You are speaking from SwiftMQ, I guess. SwiftMQ differs between system and regular queues. They differ in the queue name. System queues contain a '$' sign which is not allowed for regular queues, thus a JMS client cannot address it. Thus it is intended. It has nothing to do with the JMS spec, since you never see a system queue from a JMS client. You only get notice of it through SwiftMQ's admin tools (you see them in the SwiftMQ Explorer, for example).
A routing queue is such a system queue, prefixed with "rt$<router>". Access is forbidden for JMS clients. However, you can view and delete the messages via SwiftMQ Explorer/CLI:
http://www.swiftmq.com/products/router/admin/explorer/index.html
However, this is intended for admins.
You may ask SwiftMQ related questions in SwiftMQ's jGuru forum:
http://www.jguru.com/forums/home.jsp?topic=SwiftMQ
-- Andreas

Urgent JMS issue with SSL-enabled cluster

Hello, dear All!
We have deployed a SAP WebAS SP13 SSL-enabled cluster (2 servers) and face the following strange behaviour:
When both servers are running our queue-based message driven beans (MDB EJBs) never get any messages.
However, JMS topic subscriber threads (not implemented as MDBs) work fine on both servers and receive JMS broadcasts. As well web-initiated JMS queue browsing works fine.
Then if only one (central) server is up, queue-based MDBs work fine and start receiving messages...
If you know or guess what might be an issue it would be greatly appreciated!
Thank you and best regards,
-Yuri

Hi!
Yes, I solved this problem. You have to set your certificate to the LDAP server and get SSL enabled. You should also add same certificate to your jdk's cacerts file. That should help. :)
Janne

Queue Reference on a DVR?

Hi All,
I am trying to track down a bug in my code to no avail (so far). I wanted to check if my problem is caused by a 'feature' in LabVIEW while I continue to tear (whats left of) my hair out.....
Here is what I'm doing:
I create my 'communication queues' in my toplevel vi. These are unnamed. My toplevel vi is a queue-based state machine. I'm using OO by reference. Therefore the place where I've stored my reference is Object>DVR>Cluster>queue reference.
Using the toplevel.vi I kick off a user interface. I do this by using Static VI reference>SetVal to pass the object/DVR value to the UI and then >Run (wait til done false) to start the UI.
And here's the problem: when put an element on the toplevel state-machine queue using the UI its not received by the toplevel state machine.
Here's what I've looked at to try and find out what the problem is
The element is placed in the queue with no errors
The references are valid
One this I did notice is that when I probed the reference hex numbers they are different - my understanding was creating a new queue reference by name would give two different references to the same queue in memory (and so they have to be closed). But in my case I only create one reference to the queue, put it in my DVR and remove it as needed.......so what are the values different?
I've used this same architecture already in this project but in this case I didn't put the queue references on the DVR - instead I opened new references by name whenever I communicated between modules. This worked fine......but I'd prefer not to restrucutre my code if I can avoid it!
If anyone can offer suggestions that I can check or indeed if this is a 'feature' which has come up before I'd appreciate any comments,
Thanks,
Dave
Solved!
Go to Solution.

Hi All,
Please disregard the comments about the reference hex numbers being different - I built a very simple test program to make sure performance was correct and found that it was. In my test code both reference numbers are identical as should be expected! Its a common/garden bug causing my problem not a LabVIEW 'feature'......just gotta find the bugspray now!
Dave

Any utili to clean the messages in the queue?

Is there any utility that can delete specific messages from the queue? that could make maintenance easier.
          Regards
          dt

Is there any utility that can delete specific messages from the queue? that could make maintenance easier.                    (1) You could write a program that uses a javax.jms.QueueBrowser. Then delete specific messages by consuming them using a selector with the message identifier as in the following example:
          String selector =
          "JMSMessageID = '" + message.getMessageID() + "'";
          Keep in mind that the queue browser is a not a "live" view of the queue. It is a snap-shot.
          (2) WL 9.0 and later has message management mbeans (JMX) that are available both programmatically and through the console. These let you browse and delete messages, and are more powerful than the QueueBrowser (there's access to messages that normally wouldn't be visible to the QueueBrowser, such as scheduled messages that haven't been born yet).
          (3) There are various utilities that use QueueBrowser.
          The one I recall is a dev2dev project based on the JMS "QueueBrowser" API that's called "JMS Error Destination Manager". I think there's also a fairly sophisticated open source project (on Apache?) which works similarly (I don't recall the name).
          Tom

MQseries Queue depth

Hello,
I am using JMS for accessing MQseries. Can anyone tell me asto how can i find the current depth and maximum depth of a queue using the JMS calls.? I am using IBM java classes for JMS.my MQseries verion is 5.2
Thanks
Joe Cherian

I'm afraid that the JMS spec, for some bizarre reason, do not have any APIs for getting the depth of a queue. So, the only way to find the depth of a queue in JMS is to create a browser and browse every message in the queue, and count them. This is quite a neat solution as long as you only have a few messages. If you have thousands you will spend all your time browsing queues not processing them.
Also I believe there is also a bug in MQ 5.2 and JMS whereby if you do not browse ALL the messages, but stop halfway through instead, then you will not release some resource that MQ uses. After xx number of doing this ( depends on your maximum open handles for the queue manager ) MQ dies. So if you ARE going to browse an MQ 5.2 queue, browse all of it.
The way around it we are using is to have JMS adapters for our app, and having mutant/hybrid MQ adapters as well, which use JMS for nearly everything, including transactional gets and puts, and has an MQ queue manager for getting the MQ Series queue depth natively. Not ideal, but it allows us to use JCA/XA etc with MQ Series. I have no idea how you get the max depth of a queue, unless you can find it in the api docs then you may have to use a native call.

Cannot receive message on temporary queue

I am executing the following code in a web application.
          It is sending a message to a JMS queue. The MDB
          listening to that queue processes the message
          and posts a reply to the temporary queue specified in
          the JMS Reply to field. The queue browser code below
          shows that the message is in the queue, but the receive
          times out and returns null. My question is why can't the
          receiver get the message?
          Am I missing something?
          Thanks in advance,
          -Doug.
          =====WEBAPP SIDE===========
          QueueConnection connection = null;
          QueueSession session = null;
          QueueSender sender = null;
          QueueReceiver receiver = null;
          Queue respQueue = null;
          try
          connection = QUEUE_CON_FACTORY.createQueueConnection();
          connection.start();
          session = connection.createQueueSession(false, Session.CLIENT_ACKNOWLEDGE);
          TextMessage textMessage = session.createTextMessage();
          textMessage.setText(xmlInput);
          if (wait)
          respQueue = session.createTemporaryQueue();
          textMessage.setJMSReplyTo(respQueue);
          receiver = session.createReceiver(respQueue);
          sender = session.createSender(INPUT_QUEUE);
          sender.send(textMessage);
          if (wait)
          QueueBrowser qB = session.createBrowser(respQueue);
          Enumeration enum = qB.getEnumeration();
          while (enum.hasMoreElements())
          Message msg = (Message) enum.nextElement();
          LOG.debug("msg = " + msg);
          return ((TextMessage) receiver.receive(10000)).getText();
          return DEFAULT_RESPONSE;
          =======MDB side============
          QueueConnection queueConnection = QUEUE_CON_FACTORY.createQueueConnection();
          queueConnection.start();
          QueueSession queueSession = queueConnection.createQueueSession(false, Session.AUTO_ACKNOWLEDGE);
          TextMessage textMsg = queueSession.createTextMessage();
          textMsg.setText(buffer.toString());
          QueueSender queueSender = queueSession.createSender((Queue)getReplyDestination());
          queueSender.send(textMsg);
          queueSender.close();
          QueueBrowser qB = queueSession.createBrowser((Queue)bpelCtx.getReplyDestination());
          Enumeration enum = qB.getEnumeration();
          while (enum.hasMoreElements())
          Message msg = (Message) enum.nextElement();
          LOG.debug("msg = " + msg);
          queueSession.close();
          queueConnection.close();
          ==== LOG ====
          [DEBUG] 2006-01-16 14:51:17,780 MDB - msg = TextMessage[ID:N<192225.1137444677780.0>, <unique-id-response xmlns
          ="htt...]
          [DEBUG] 2006-01-16 14:51:27,374 WEBAPP -
          msg = TextMessage[ID:N<192225.1137444677780.0>, <unique-id-response xmlns="htt..
          .]

At first glance, this looks like a standard design to me. Does the same problem occur when there are no browsers?
          Tom
          (Performance note: It is quite CPU-intensive to continually create and close connections/temporary destinations/sessions/consumers/producers. Cache and/or pool them for re-use when possible. Use a cached anonymous producer on the MDB for sending replies [pass null for the dest when creating it, and specify the dest as a send parameter].)

OC4J/JMS Queue Listener

Hello,
Have a simple test JMS client that is listening for messages posted to a JMS queue setup in the OC4J/JMS. While the messages are ending up in the proper queues [this can be verified via a queue browser] the listener does not seem to be getting any notifications. The receiver code is essentially,
queueConnection = queueConnectionFactory.createQueueConnection();
QueueSession queueSession = queueConnection.createQueueSession(
                         false,
                         Session.AUTO_ACKNOWLEDGE);
QueueReceiver queueReceiver = queueSession.createReceiver(queue);
System.out.println("Set Listener");
queueReceiver.setMessageListener(new OracleListener());
queueConnection.start();
This is from a remote client for both the sender and the reciever. Any help would be much appreciated.
Vinay Menon

I am investigating this issue myself.   I have come up with two possible scenarios.
1. Do it in Java -- Of course this defeats the purpose of using SOA Suite. In the book SOA Suite 11g handbook, chapter twelve he has a jms adapter publishing to a queue and a java program listening to that queue.    http://technology.amis.nl/2010/09/21/soa-suite-11g-handbook-chapter-complements/
you can refer to the TheFinancialApplication.
2. If you have access to the Jms programs that publish the messages to the queue (that your programs want to listen to) and they are done in SOA Suite 11g, you could use Event Delivery Network (EDN) to create a event on the publishing side-- and then set up your listening processes to listen for that Event (which is jms under the hood).
You could then have those processes kick off from that event.
Anyone with a better idea?
-- Update 8 hours later -- I have a better idea:
I created a composite that publishes a JMS message to a queue on weblogic using the jms adapter. I created a second composite with a jms adapter (on the Exposed services side of the composite) that Consumes a message, hooked it up to a mediator and then wrote the contents of the message to a file. Uploaded both.
Moments after the JMS publisher Composite ran, a file was created. From this I would surmise that you could run other processes as well.
-- Update two days later.   So much for the better idea.
Given that the Flow Trace of the last test looked suspiciously like this one one application and not two separate applications, I decided to test it further. I have two PCs. On each PC I have a soa domain, Uploaded the jms publisher to one domain, and the JMS Consumer to the other separate domain on the other PC.   Ran the publisher, and the other one just sat still -- didn't start at all.
The first test was too good to be true!
Stuart

Cannot connect to cluster

In Crystal Reports Server XI release 2, I have the Servers running correctly, and connected to the correct database, I have all of the services running using the correct clustername. However I am getting this error when I try to run a report.
"Unable to connect to cluster @myclustername to retrieve updated CMS member list. Logon cannot continue."
Can anyone assist with this?
Thanks

Hi,
Just to be clear, You have a Cluster running and you can connect to the cluster on one Hyper-v node ?
And if you looked at this cluster do you see the other nodes ? if not no problem. If you do see the nodes check the state of this nodes.
Open the Failover Cluster Manager on the first node and connect to the cluster / do browse of add a " . "
Check this blog post this could be helpful
http://robertsmit.wordpress.com/2013/06/12/windows-server-2012-configure-deploy-hyper-v-replica-broker-how-to-tee13-msteched/
Greetings, Robert Smit Follow me @clustermvp http://robertsmit.wordpress.com/ “Please click "Vote As Helpful" if it is helpful for you and Proposed As Answer” Please remember to click “Mark as Answer” on the post that helps you

Consuming from Queue

I am working with a group, and they need to be able to look at an item on the Queue without consuming it. I think this can be done?? The item you are looking at should be locked to keep other elements from using it while you evaluate this using the item.
If that makes sense???
I am not sure if the default JMS implements this behavior. Can anyone confirm..

Hi,
Yes, you can browse messages from a queue (have a look at QueueBrowser). However a queue browser returns a snapshot therefore the messages that are browsed are not �locked� i.e. they can be consumed and browsed.
A solution that may be useful to you is to consume the queue messages within a transacted session. The consumed messages are not visible to other consumers and your client application can have a look at them in isolation. You can then rollback the transaction whenever you need to �release� those messages. Note that if you don�t want to reinsert those messages in the queue you only need to commit your transaction. It is a bit of a hack but it should work.
Hope this helps,
Arnaud
www.arjuna.com

JMS Connectivity problem

Hi,
On one of our production deployments using Weblogic 10.3.0, we are getting the following exception when trying to connect to a JMS queue on one of the managed server. We are able to access the queues on other managed servers(there are 4 managed servers in all).
javax.naming.NameNotFoundException: Unable to resolve 'weblogic.messaging.dispatcher.S:+<ManagedServerName>+'. Resolved 'weblogic.messaging.dispatcher'; remaining name 'S+:<ManagedServerName>+'
This error message starts coming up after the server has been up for few days. Once this exception comes up, we cannot fetch messages from the JMS queues on the affected managed server. We have a java process that connects to the managed servers and picks up the JMS messages and sends it to external systems. The problem occurs when connecting from this process. I tried to find the queue depth by writing a simple java program that uses a JMS queue browser. The program fails too.
I checked the JNDI tree to find if a binding exists with 'weblogic.messaging.dispatcher.S:+<ManagedServerName>+'. To my surprise, the binding exists. Just wondering what could possibly cause this issue.
My JMS settings are like follows:
I have four JMS servers, each targeted to the corresponding managed server.
I have distributed JMS queues that are targetted onto the cluster.
I have normal JMS queues that are targeted onto the corresponding managed server.
The connection factory is targeted onto the cluster.
The problem happens with my normal JMS queues on one of the managed servers.
Please let me know if I am going wrong with my setup. Note that this issue happens after the server is running for few days. Upon restarting the affected server, I am able to connect to the JMS queues and pick-up messages.
regards,
Anand
Edited by: Anand1982 on Jun 5, 2011 10:51 PM

I can think of two potential reasons why this error might come up. The first is that your WebLogic Cluster might be failing so that JNDI name-space replication between cluster servers has at least partially stopped working. The second is that you're working with multiple domains but there are duplicated names configured between domains. For the former, look for Warnings in your logs and also consult the general documentation for cluster troubleshooting information. As for the latter, consult the best practices section of the JMS configuration guide.
Regards,
Tom

Clustering across subnets.

I am trying to split our CF multiserver cluster into seperate
subnets, and I am having great difficult making it work. I am
hoping that someone else here has done this, can can be of
assistance.
Background:
2 boxes both running mx7 multiserver EE.
2 Jrun servers each box, one cluster.
IIS on both boxes pointed to cluster. Browsing successful
from both webservers.
System was 100% operational pre-split. Testing confirms all 4
severs processing CF code in RR fashion.
Changes:
Moved one server to seperate subnet seperated only by a SOHO
router.
SOHO router configed in router mode, "inside" server set as
DMZ host for 100% port forwarding (no rules)
jrun servers deleted and re-built w/ ip addresses instead of
DNS names.
edited security.propertes to update jrun.trusted.hosts
Results of move:
host 1 - cluster controller operates as expected, less 2
remote jrun servers
2 remote servers as show as stopped.
can access jrun admin and jrun servers cf admin pages (8301,
8302) on remote host
host 2 (remote) - WSconfig can see only default "cfusion"
server.
wsconfig unable to see cluster or other jrun servers on host
1.
IIS will process pages if connected to host 1 - "cfusion"
can access jrun admin and jrun servers cf admin pages (8301,
8302) on remote host

here is the gospal on clusters although I'm not sure it will answer your question
http://download.oracle.com/docs/cd/E10291_01/core.1013/e10294/toc.htm
what do you mean by subnet, generally all machines are on the same subnet as they sit side by side?
cheers
James

Not able to Read All the Message in the Topic

Hi Al,
i am new to JMS. I have implemented publish-subscribe model. Topic as a destination.
I want to read all the message available for particular topic. is there any way to get this?..
like in the Queue, QueueBorrower used to read all available messages in the queue.
Thanks
Raj

It depends on what you mean by "what messages are available on the topic".
if you want to know what messages would be delivered to a new consumer on the topic, then you can simply create a consumer yourself. Each messages on a topic is delivered to all subscribers, so by creating a consumer yourself you won't be interfering with the delivery of messages to other consumers. (That's not possible with queues which is why the queue browser is provided).
If you want to know what messages are "persisted in the topic" then there's not really such a concept. Messages sent to a topic are only persisted in individual durable subscriptions created for individual clients. If there are no durable subscriptions when messages are not persisted. The JMS APi doesn't provide a way to browse the messages saved in an individual durable subscription. It could, but it doesn't.
Nigel

Glassfish3 cluster queue browser

Similar Messages

Maybe you are looking for