ProxyFactory#getProxy seems to re-use EVM instances from the cache without properly settings optionalParameters about blaze-persistence

[Mobe91]

@EugenMayer would you be able to try if the changes in #1806 solve this?

[Mobe91]

Thank you for the hints. If you are curious, I've opened a spring-boot issue a while ago for this and included quite some details about what I am observing in my case. Reducing the application context cache size was the first thing I did but strangely the application contexts are still not garbage collected even though they are no longer held by the cache.

My tests run in Maven but no parallelism is involved. I might give the EM clearing a try though! 👍

[EugenMayer]

@Mobe91 thank you a lot for working on this item! Will try to queue the test today

[EugenMayer]

@Mobe91 indeed this solves my issue entirely. Did it solve your memory leak issue too?

[Mobe91]

Did it solve your memory leak issue too?

Unfortunately not, there must be other things going on in my case that are probably not related to blaze-persistence 🥲

Glad to hear that it works for you! @beikov will make sure it goes into the upcoming release.

[EugenMayer]

@Mobe91 AFAIK you are using spring boot in your tests right?

Some aspects we use:

1. I'am sure you are aware about the context cache and the ability to limit the amount of caches spring boot is preserving during the tests? For us, this has limitted the used memory and connections. spring.test.context.cache.maxSize=5 or similar helps our cause.

2. In addition, you might want to implement an TestExecutionListener that clears the EM L1 cache after every test or test suite.

3. Also, if you are using gradle you might tune your test task

    maxParallelForks = 1
    systemProperties['junit.jupiter.execution.parallel.enabled'] = false

We use this for all JPA based tests.