This article indicates that if MQ has to be used as a Transaction Manager, the application has to compulsorily use Bindings mode.
http://www.ibm.com/developerworks/websphere/library/techarticles/0601_ritchie/0601_ritchie.html
Bindings mode also means that the MQ Server and the application both have to be on the same host because it uses JNI.
My question is, what is the fundamental reason behind this requirement? Why can't MQ Manager work like JBOSS which can reside on any host and take requests for managing distributed transactions?
What prevents WMQ from supporting client connection for distributed transactions?
Thanks,
Yash
That article is very old, it was from 2006, for a long time now MQ clients that are located remote from a queue manager have been able to use XA transactions.
MQ supports XA with JMS when using a supported transaction manager within a JEE container
http://www.ibm.com/support/knowledgecenter/SSFKSJ_8.0.0/com.ibm.mq.dev.doc/q032590_.htm?lang=en
Related
We are designing a solution that will consume messages from IBM MQ using JMS. The plan is to use WAS Liberty, so JMS is the technology of choice. We will create Message-Drive beans that will listen for messages in MQ queues.
We are considering both WAS Liberty and OpenLiberty as well.
The trick here is that we must implement it with fail-over, so that if one of our server fail, the other will keep consuming messages from MQ automatically. Like in a ative/passive mechanism.
I'm aware that the MQ adapter needs to be installed as it is not provided out-of-the-box.
I have the following questions:
Does WAS Liberty messaging implementation supports fail-over? Meaning that if the ative message consumer node fails, the stand-by node will automatically migrate and start consuming messages from MQ? What about OpenLiberty?
How can I configure the message system to work that way? Can you point out to the documentation?
Or is this feature only provided by WebSphere?
There is no such functionality in WebSphere Liberty or Open Liberty yet. You can create RFE here https://www.ibm.com/developerworks/rfe/?PROD_ID=544 .
There are ways to do it manually, check these links:
JMS Activation spec on Liberty: “WAS_EndpointInitialState” full profile equivalent property?
Controlling the state of endpoints at runtime
Solution that you could do:
create a script/application that will monitor your servers and call that API to enable/disable endpoint in specific server
or use Dynamic cluster/ auto scaling feature of Liberty and divide you app to two clusters - one with MDBs, one without. And then define policy that MDBs cluster has 1 instance always available. So once the server dies it is automatically restarted somewhere in the cluster
or use Kubernetes/ICP platform in the same way - so deploying 2 versions of app, and defining different replicasets parameters.
I am not expert in JMS and recently came across a situation in my project where they wanted to use JMS to communicate between applications. Is it possible to make use of JMS without an underlying MQ software like Websphere MQ / TIBCO EMS to communicate asynchronously. My thinking is that just like we use JDBC API to speak to a physical Database product (like Oracle/DB2 DB) underneath, we would use JMS to speak to a physical queue. But my team says just JMS and no queue. Can anyone help me understand? We use Websphere 6.0 server and use spring framework.
Thanks in advance.
You are correct: JMS is solely an interface, it contains no implementation. The comparison with JDBC is quite fitting.
When using JMS you always need a JMS provider. A provider is the JMS implementation, which interfaces a message oriented middleware.
Most application servers have built in JMS providers, so maybe your team meant using JMS without an external provider.
Is it possible to make use of JMS without an underlying MQ software like Websphere MQ / TIBCO EMS to communicate asynchronously
This makes no sense. Few points to note -
JMS is just specs. By specs you can imagine it as set of APIs/interfaces/ method signatures and governing rules.
There are various organization/companies that provide implementation of these APIs.Apache's ActiveMQ,IBM's Websphere MQ or TIBCO EMS are some of the examples.
Now when you say I want to use JMS then you need a JMS compatible server i.e server which understands the rules laid down by specs. This is generally know as MQ server and is provided by the provider.
My thinking is that just like we use JDBC API to speak to a physical Database product (like Oracle/DB2 DB) underneath, we would use JMS to speak to a physical queue. But my team says just JMS and no queue
Two things to note here. When you say you want to use databases and JDBC APIs to communicate JMS comes no where in the picture. Please understand why JMS is used. In short it is used to transfer messages between two decoupled system. What storage type JMS server uses(and if it is configurable) will depend on the provider. Generally all have DB storage type and can be configured to use your DB.
Next JMS has two types of communications - PTP(peer to peer) which uses queues and PUBSUB(publish subscribe) which uses Topics. When you say you are using JMS everything boils down to either of these two and some of it's variants.Now when you say JMS without a queue perhaps your team meant using PUBSUB. But again it is not something you decide randomly and use it in your application. This decision is takes as per what your requirement is as both of them behave differently.
First you will need queues/topics to use JMS.
Second (your team probably means not using WebSphere MQ because my team was saying the same thing)
This diagram in IBM's Red Book will fully answer your question (it is the same as WS6):
http://www.redbooks.ibm.com/redbooks/pdfs/sg247770.pdf
page 4.
Probably what you will be using is the WebSphere Default Messaging Provider.
An earlier question on interoperating with MQ Series and a pure open-source solution was answered partly with:
"The one exception is that WebSphere App Server Messaging Engines can interoperate with WMQ. These are pure JMS messaging engines written in Java but they understand the WMQ formats and protocols and appear to WMQ as another QMgr. So if you have WAS you can talk to WMQ without a WMQ client or another QMgr."
(http://stackoverflow.com/questions/18236860/can-we-talk-to-remote-websphere-mqseries-purely-with-non-websphere-mqseries-soft)
The documentation I've read through says that
WebSphere Application Server using WebSphere MQ link can operate with a service integration messaging engine or queue-sharing group.
So as I understand it, if you have WAS and WebSphere MQ link you could connect an WebSphere esb to an exterior MQ Series QMgr and look like a QMgr to that exterior MQ Series QMgr.
Without a WebSphere esb though can WAS act like a MQ QMgr and somehow manage third-party JMS messaging providers like ActiveMQ?
Thanks for any help
I believe you are mixing up capabilities.
WAS can absolutely appear to WMQ as another QMgr when using the WMQ Link. The QMgr appears to WAS in this configuration as another Service Integration Bus.
Service Integration Bus or SIB is the WAS abstraction for a JMS transport provider. It isn't the same thing as IBM's Enterprise Service Bus (ESB) product. Both talk to WMQ but WAS is in no way dependent on the ESB product to do so.
Either WMQ or any other pure JMS transport can be configured in WAS as a Foreign JMS Provider. The administrator points WAS to the JMS classes provided by the transport provider. These can include Active MQ.
There is no concept of WAS seeing WMQ but managing some other transport such as Active MQ. WAS can use normal JMS API calls to interact with Active MQ or any other transport provider. It can also make use of whatever administration API is provided by that transport provider using that provider's administration API. Since both WAS and MQ are IBM products, the integration is a bit tighter between them which is why you can use WMQ natively whereas other JMS transports must be configured as foreign JMS providers.
I am developing a service for JMS messaging using IBM WMQ version 7.5 as Message oriented middleware. I am in a dilemma. The IBM specification says to use WebSphere MQ JMS API in JAVA EE environment. My sender and receivers will be in remote JBoss deployments. IBM provides a resource adapter to connect via JCA. This connection requires static administration of components such as Queue managers, queue channels, queue names.
Is JCA the best I can do ? Is there any other way where I can dynamically create the queues provided that I know other configuration details ? There will be multiple applications using this setup running on their own servers. If I use JCA, all have to make their own configurations. Can I make dynamic configurations ?
Typically within the application server environment resources would be created in some JNDI context. These would be Connection Factories and Destinations.
Depending on the app server, it may be possible to share this configuration across a network of individual servers. So typically there would be some set of static definitions.
However within say a servlet the MQConnectionFactory can be dynamically created.
In all cases the backing WMQ Resources do need to exist; though it is possible to send administrative command messages to WMQ to create queues and do other admin. The only exception is topics, accessing a topic string brings it into existence. Though this has default security properties that you might not want.
I understand JMS spec allows me to write MDBs. Then I can deploy them on an appserver like websphere or jboss (and they have inbuilt implementation of JMS provider which is Java EE compliant). The question is when to use a full fledged product like IBM WebSphere MQ or ACtiveMQ instead of using App server's inbuilt JMS provider.
This appears to be a good reference to what you are asking, specific to IBM products but can be generalized to using an application server integrated messaging system vs a standalone messaging system - http://www.ibm.com/developerworks/websphere/library/techarticles/1109_wallis/1109_wallis.html
A built in JMS provider, such as WebSphere SI Bus, possibly others, are built for JMS messaging.
WebSphere MQ (MQSeries), ActiveMQ etc. have clients for multiple languages and are built for applications other than Application Server hosted applications to connect to it (C++, Java SE, Python, etc).
Say you want a local messaging server on the same physical server (or network?) as a non JavaEE application, you don't want to install a Java EE App Server, but want a stand alone messaging server instead (WebSphere MQ, Active MQ).
That said, ActiveMQ, for instance, is the messaging system inside Apache Geronimo. So there is overlapping here.
Then there is a whole discussion about management, clustering and security. Java EE messaging systems are pretty much optimized to follow the cluster layout of Java EE. In other deployment scenarios, the clustering might be optimal in other ways, such as cluster for different security zones. Also security models and management tools might be different in stand alone messagaing servers.