Disable caching in JPA (eclipselink)

Question

I want to use JPA (eclipselink) to get data from my database. The database is changed by a number of other sources and I therefore want to go back to the database for every find I execute. I have read a number of posts on disabling the cache but this does not seem to be working. Any ideas?

I am trying to execute the following code:

        EntityManagerFactory entityManagerFactory =  Persistence.createEntityManagerFactory("default");
        EntityManager em = entityManagerFactory.createEntityManager();

        MyLocation one = em.createNamedQuery("MyLocation.findMyLoc").getResultList().get(0);

        MyLocation two = em.createNamedQuery("MyLocation.findMyLoc").getResultList().get(0);    

        System.out.println(one==two);

one==two is true while I want it to be false.

I have tried adding each/all the following to my persistence.xml

<property name="eclipselink.cache.shared.default" value="false"/>
<property name="eclipselink.cache.size.default" value="0"/>
<property name="eclipselink.cache.type.default" value="None"/>

I have also tried adding the @Cache annotation to the Entity itself:

@Cache(
  type=CacheType.NONE, // Cache nothing
  expiry=0,
  alwaysRefresh=true
)

Am I misunderstanding something?

Fingernail

James in your comment to my answer, was the caching off ( <property name="eclipselink.cache.shared.default" value="false"/> ) when you tested it? – Locust

Sorry just noticed this, yes the caching was off. I am still having this issue and am no closer to a solution. – Fingernail

Answer 1

This behavior is correct, otherwise if you change object one and object two with different values you will have problems when persisting them. What is happening is the call to load object two updates the entity loaded in the first call. They must point to the same object since they ARE the same object. This ensures that dirty data cannot be written.

If you call em.clear() between the two calls, entity one should become detached your check will return false. There is however no need to do that, eclipse link is infact updating your data to the latest which I would guess is what you want since it frequently changes.

On a side note if you wish to update this data using JPA you will need to be obtaining pessimistic locks on the Entity so that the underlying data cannot change in the DB.

You will need to disable the query cache as well your cache options were just removing the object cache from play not the query cache, that is why you are not getting the new results:

In your code:

em.createNamedQuery("MyLocation.findMyLoc").setHint(QueryHints.CACHE_USAGE, CacheUsage.DoNotCheckCache).getResultList().get(0);

Or in persistence.xml:

<property name="eclipselink.query-results-cache" value="false"/>

Locust

That helps thank you. I will have to look at pessimistic locks as the data will be changing. How then if I add a Thread.sleep(10000) between the queries and in that time manually change an attribute of MyLocation in the database does two (and therefore one) not reflect this change? – Fingernail

added a link to some info about pessimistic locks – Locust

I have tried the described change in the code and persistence.xml and still don't get the value I changed directly. Any ideas? – Fingernail

Have a look at your transaction isolation levels. If your transactions use Serializable as isolation levels (default in Oracle for example), than T1 will never see changes from T2 (AFAIK). – Grindle

DoNotCheckCache will make sure the cache won't be cached. Is it possible to make sure that the returned entity by that query is not cached ? – Jaquelin

@AliReza19330 This answer is 6 years old now, feel free to update it if it no longer works. there are however still upvotes taking place now and then. – Locust

@justin Yes. sure, I think the fault might be related to the version of technologies and the jpa implementation, then I post the solution which works for me as a new answer. – Impracticable

Answer 2

final Query readQuery = this.entityManager.createQuery(selectQuery);
readQuery.setParameter(paramA, valueA);

// Update the JPA session cache with objects that the query returns.
// Hence the entity objects in the returned collection always updated.
readQuery.setHint(QueryHints.REFRESH, HintValues.TRUE);

entityList = readQuery.getResultList();

This works for me.

Arnettaarnette

This was THE solution for my problem; I am bound to JTA, and inside JTA you're not allowed to open Transactions. No Transaction, no refresh(). This solution here solved my problem. – Pasha

Answer 3

If you wish to disable caching without getting vendor specific, you could annotate your domain object with:

@Cacheable(false)

Here is an example:

@Entity
@Table(name="SomeEntity")
@Cacheable(false)
public class SomeEntity {
    // ...
}

Thar

This should be definitely the valid answer ! – Panegyrize

God bless you, sir. This issue has been driving me nuts and you managed to solve it for me in 5 seconds. I wish I had more mod points. – Okeefe

This is definitely the best answer here. – Mathia

Answer 4

See,

http://wiki.eclipse.org/EclipseLink/UserGuide/JPA/Basic_JPA_Development/Caching

For the same EntityManager JPA always requires that one==two, so this is correct, not matter your caching options (this is the L1 cache, or transactional cache, which enforces your transaction isolation and maintains object identity).

To force the query to refresh (and revert any changes you have made) you can use the query hint "eclipselink.refresh"="true". Or probably better, use a new EntityManager for each query/request, or call clear() on your EntityManager.

<property name="eclipselink.cache.shared.default" value="false"/>

Is the correct way to disable the shared cache (L2 cache). Please remove all your other settings as they are not correct, and can cause issues.

EclipseLink does not maintain a query cache by default, so those settings will have no affect. CacheUsage is also not correct, do not use this (it is for in-memory querying).

Fingernail

Note that "eclipselink.refresh"="true" is buggy in current versions of EclipseLink (bugs.eclipse.org/bugs/show_bug.cgi?id=398074); also, it is EclipseLink-specific and not part of JPA. So I'd advise to just get a fresh EntityManager whenever you want fresh data - that's how it is supposed to work with JPA. – Bridie

Getting a fresh EntityManager will not get you a fresh object directly from the database. The cache work on a EntityManagerFactory level, which normally has the life time of the application. – Rayburn

Answer 5

First level cache is enabled by default and you can not disable it. i.e. no settings in your persistence.xml file will disable first level cache.

You can only clear out all the entity manager objects by calling

entityManager.clear()

this will make subsequent queries go to the database (the first time) and then objects are again stored in the cache

You can force each query to go to the database directly by calling

query.setHint("javax.persistence.cache.storeMode", CacheStoreMode.REFRESH);

Blackmarket

Answer 6

I know this post might be old, but I am writing for others who need help. I had this problem and finally I solved it by this code:

em.createNamedQuery("findAll").setHint(QueryHints.CACHE_RETRIEVE_MODE, CacheRetrieveMode.BYPASS).getResultList();

It works really well. And we can see in javadoc of the BYPASS enum, it is written that:

Bypass the cache: get data directly from the database.

Note that I use Weblogic 12c and TopLink as a JPA implementation.

Impracticable

The refresh hint mentioned above did not work for you? – Otti

I didn't try that, because I wanted to bypass jpa cache just in some queries. – Impracticable

Answer 7

If manually modifying the attributes of the object, for example MyLocation. The above trick (CACHE_USAGE=CacheUsage.DoNotCheckCache, or eclipselink.query-results-cache=false) does not seem to work as I tried.

So i tried to set another hint which is eclipselink.refresh, to true. then it works. I mean the manually changed attributes get retrieved.

So as i understand, the above trick only ensure the it gets the correct objects. However, if the objects have been cached already, eclipselink just returns them without checking the freshness of the contents of the objects. Only when the hint eclipselink.refresh is set to true, will these objects get refreshed to reflect the latest attribute values.

Pathway

Hava you set this in your persistence.xml? If I add this hint to the query it works. But it does not work if I define it only in my persistence.xml. – Nahuatlan

Answer 8

When you create your query via the entity manager the first-level cache is called.The entity will be old. Lets say I have a key ,which I updated manually in the database and then tried to retrieve:

Key result = entityManager
                .createQuery("SELECT k FROM Key k WHERE k.linkKey = :key", Key.class)
                .setParameter("key", key)
                .getSingleResult();

The following key will be as it was ,because it is still cached ,my update will not appear.To refresh the entity you will need to call this:

 entityManager.refresh(result);

Naidanaiditch