Java WebApp choosing JTA, database replicator/load-balancer or both - postgresql

We have a webapp that is currently running on one instance of Apache Tomcat with one database instance, but the increase in traffic will soon (probably) force us to resort to load-balancing several webapp instances, and we've run into a problem that seems to have no easy answer.
Currently our JDBC DataSource is configured as Resource-local, rather than Transactional, and after some searching, everyone recommends to use Transactional, which requires the use of a JTA provider. No real justification is used for why I don't just stick with the current scenario where we have a servlet filter catch any unhandled exceptions and rollback the active transaction. Besides that the only one I've found that is just a JTA provider (not with 5 more JEE technologies combined) and is still maintained is Bitronix. The other alternative is to move out of Tomcat and use Glassfish, since it is a full Java EE platform, and we also use JavaMail, JPA and JAX-RS.
Only one transaction scenario uses Serializable isolation level.
As for the database, we may be looking too far ahead to think of distributed storage like Postgres-XL or pgpool, but if we make the wrong choice now it will be harder to fix later.
My questions are as follows:
Do synchronous database replication tools and JTA complete each-other, hinder each-other or just perform the same consistency checks twice?
Do we need JTA if we only have one database, but multiple webapp instances?
Do we need JTA if we have multiple database and multiple webapp instances?
Should we just switch to Glassfish or something like TomEE?
Supposedly there are ways we can keep using Hibernate as our JPA under both. It would be tedious to have to rewrite all our native queries to use positional parameters because EclipseLink and OpenJPA don't support them. That little extra feature makes Hibernate worth choosing above all other JPAs for me.

Related

Spring boot Kogito Mongodb integration

I'm working on creating a kogito bpm spring boot project. I'm very happy to see reduced level of complexity on integration on jbpm in spring boot with the help of KOGITO. I'm struggling to find answers for my question, so posting them here,
Kogito is a open source cloud offering for jbpm. I'm I correct?
I see only mongodb or infinispan can only be used or supported with Kogito. I can't integrate psql with kogito. I'm I correct?
I successfully created the spring boot kogito mongodb project and when I placed a .bpmn file in the resource folder, automatically endpoints got created. I was able to access, run the process and get response. But I don't see any entries created in MONGODB :| I don't even see the table being created. The .bpmn contains a simple hello world flow with start+scripttask+end nodes. Please explain help me understand this. Is the RuntimeMangar configured for per request stratergy? How can I change it?
Answers inline.
Kogito is a open source cloud offering for jbpm. I'm I correct?
Kogito is open-source and has jBPM integrated into its codebase to run on a cloud-native environment. In addition, a lot has been made to make it also run with native compilation if used with Quarkus.
I see only mongodb or infinispan can only be used or supported with Kogito. I can't integrate psql with kogito. I'm I correct?
To this date, Kogito has the following add-ons to support persistence:
Infinispan
Postgres
MongoDB
JDBC (so you can extend to support any database you wish)
See more about it here https://docs.jboss.org/kogito/release/latest/html_single/#con-persistence_kogito-developing-process-services.
But I don't see any entries created in MONGODB
Do you mind sharing a reproducer? Have you taken a look at the examples in https://github.com/kiegroup/kogito-examples/tree/stable/process-mongodb-persistence-springboot? This example shows a call to a sub-process that relies on a user task. Hence the process must be persisted to fire up again on a new request to solve the task. However, since your process starts and ends in one request, there's nothing to be persisted in the DB:
Runtime persistence is intended primarily for storing data that is required to resume workflow execution for a particular process instance. Persistence applies to both public and private processes that are not yet complete. Once a process completes, persistence is no longer applied. This persistence behavior means that only the information that is required to resume execution is persisted.

CDI and User Defined Functions (UDF) in Teiid / Wildfly

Recently I began working with Teiid and Wildfly. I have a user defined function (UDF) that adds custom functionality to Teiid, and it works as expected. However, I need to modify it further and would like to use CDI to inject a bean from the Wildfly app server. I know that the UDF isn't managed by the container (it is a Wildfly module with an associated module.xml file deployed as a jar), so I've added (what seemed to be) necessary dependencies to module.xml but it doesn't work.
Is it possible to use CDI in a UDF with Teiid / Wildfly, and if so, how?
No, it is not possible. although Teiid is a resident of WildFly it is using the infrastructure of WildFly for a variety of features like transactions, security, data sources, administration etc. It is not part of JEE or anything, so there is no direct way to do this. If you want to explain what you are trying to accomplish, maybe we can offer any further guidance on alternatives.

jBPM Repositories disappear after Wildfly restart

Pardon if I can't give more pointers, but I'm really a noob at wildfly. I'm using version 9.0.2.
I have deployed jbpm-console, drools, and dashboard - no problems here. I restart wildfly using the jboss CLI, and when I login again, the repositories won't appear in the web interface or on disk (atleast nothing that grepping or find will show).
I'm using the H2 database. I'm not even sure where to look, does anyone have any idea?
Thanks in advance!
After enough reading through the docs, it would seem that it's necessary to configure jBPM to persist. From the docs:
"By default, the engine does not save runtime data persistently. This means you can use the engine completely without persistence (so not even requiring an in memory database) if necessary, for example for performance reasons, or when you would like to manage persistence yourself. It is, however, possible to configure the engine to do use persistence by configuring it to do so. This usually requires adding the necessary dependencies, configuring a datasource and creating the engine with persistence configured."
https://docs.jboss.org/jbpm/v5.3/userguide/ch.core-persistence.html

Using Cassandra and MySQL together with JPA in Play framework

I would like to use Cassandra NoSQL server with an RDBMS in Play 2.3.0!
Started to build it up using Kundera, according to this tutorial:
http://recipes4geeks.com/2013/07/06/play-nosql-building-nosql-applications-with-play-framework/
It works fine, and I can use it with pure mysql-jdbc connection, and it also works if I use jdbc for Cassandra connection and JPA for MySQL..
.. but the goal is to use a persistence framework, without handling basic JDBC stuffs!
It looks, this problem was mentioned in the link above:
Caution: javaJdbc app dependency downloads hibernate-entitymanager jar file that interferes with Kundera. Make sure you remove this app dependency which is by default present.
If I remove the hibernate-entitymanager from the dependencies, the project runs, but when it wants to call the Persistence.createEntityManagerFactory("mysql") method, Play says: No Persistence provider... as it was expected.
If I keep the hibernate-entitymanager in the dependencies list, beside the kundera client, the Play server simply shuts down.
Is there a possibility to make it work or I have to replace Kundera?
DataNucleus JPA supports persistence to all RDBMS around (via JDBC), as well as to Cassandra, MongoDB, Neo4j, LDAP, HBase and many others. It's Cassandra support seems to be for all latest versions and uses the native Cassandra driver (not JDBC) and so no chance of conflicts like above. You can read up on it at
http://www.datanucleus.org
Caution: javaJdbc app dependency downloads hibernate-entitymanager jar file that interferes with Kundera. Make sure you remove this app dependency which is by default present.
This should not be an issue with latest Kundera releases. Also you can email sample project at kundera#impetus.co.in in case looking for quick support.

Difference between using JPA with or without OSGi

I'm beginner in OSGi, My project consists of developping and executing, within an OSGi container (apache felix; the distribuable jar), a persistence bundle (using jpa) and then communicating with the database (MySql) through a jpa provider (Hibernate).
I read about the jpa specification for OSGi, so, if I have correctly understood, I must use a JPA provider for OSGi implementing the OSGI jpa enterprise specification. this jpa provider will track for a registered persistence bundle to make an EntityManagerFactory for it ?
So what is the difference between using a jpa provider directly to create the EntityManagerFactory (Persistence.createEntityManagerFactory("xx")) or retrieving it from the registry :
serviceReferences = context.getServiceReferences(
EntityManagerFactory.class.getName(),
String.format("(%s=%s)",
EntityManagerFactoryBuilder.JPA_UNIT_NAME,
persistenceUnit));
I wouldn't like to use any container (apache karaf, geronimo, spring dm, ..) so, is it sufficient that I will instal and start in the OSGi container for example the "org.apache.aries.jpa.api" as an implementation of the OGSi enterprise jpa spec and then only retrieve an "EntityManagerFactory" service from the registry associated to my persistence unit name, or I should also register by myself a PersistenceProvider like HibernatePersistence to can declare it as "provider" in my persistence.xml file ?
I found many discussion in this topic here. I still having trouble, though
Thanks
OSGi is about services and services are extremely easy to consume in OSGi with the proper setup. You show a very old style example with service references and I do agree, in that model it is a lot easier to just use the old fashioned JPA way.
However, if you use Declarative Services, using services becomes very lightweight. You get injected with an EntityManagerFactory service that is completely prepared for you. The deployer could have tuned all kinds of settings with Config Admin, connection pools, another JPA provider, etc. It is a clear separation of concerns.
By not really knowing where this thing comes from and who implements it you get less assumptions in your code and your code will therefore be less error prone and more reusable. I principle, the fact that use Hibernate and MYSQL is completely irrelevant to most of your code. Yes, I do know that neither JPA nor SQL is very portable in practice but there are many aspects that are ignorant of the differences. It is the deployer that is then in the end responsible to put together the parts that work.
Now Declarative Services (DS) is of course an extra bundle but after using OSGi for 15 years now I declare any OSGi developer NOT using DS, well, let me not go too deep into that to keep it civil :-) If I would be back in the beginning of OSGi, DS would have been built into the framework, it is for the lowest level to program with.