Spring @Transactional: interaction between isolation and propagation of transactions

Question

This directly relates to this other question, as a natural follow-up question (which was asked in comments, but was never replied to).

Textual description

Assuming:

• the application uses the default REQUIRED transactional propagation
• the application uses the DEFAULT transactional isolation
• the database's transactional isolation defaults to READ_COMMITTED (e.g. Postgres)

Two operations exist in separate classes:

1. Read some entries
2. Read some entries + Update them

Now assume that:

• the second operation calls the first one to fulfill the "Read some entries" part
• the second operation's method is marked with @Transactional(isolation=Isolation.REPEATABLE_READ)
• the first operation's class is marked with @Transactional

How does Spring behave in terms of managing transactions when calling the second operation?

---

Code description

@Service
@Transactional
class Service1 {
  @Autowired MyRepo repo;

  void read(String id) {
    repo.findById(id);
  }
}

@Service
@Transactional
class Service2 {
  @Autowired Service1 readCommittedService;
  @Autowired MyRepo repo;

  @Transactional(isolation=Isolation.REPEATABLE_READ)
  void whatHappensNow(String id) {
    readCommittedService.read(id);
    repo.modifyRow(id);
  }
}

What happens when we call service2.whatHappensNow("foo") ?

Will Spring create a transaction, with REPEATABLE_READ isolation, and use that same isolation level when calling the repository via the service1.read("foo")?

And if a concurrent request is received, but this time it's a direct call to service1.read("foo") : will that transaction have the appropriate READ_COMMITTED isolation level, thus having no chance of failing or being blocked by that other request coming from service2?