When using a sequence ID generator with MySQL Dialect, it seems to be possible to get the system to lock up completely.

This behaviour is triggered if a large number of threads are all creating entities in the table with the sequence generator at the same time.

After poring over stack traces, the problem seems to be as follows:

• The TableStructure generator is used on MySQL since there is no native support for sequences
• This generator makes an SQL call, but does this in a separate connection from the currently executing connection (see here: https://github.com/hibernate/hibernate-orm/blob/master/hibernate-core/src/main/java/org/hibernate/id/enhanced/TableGenerator.java#L530 )
• If no connection is available because other entity creation threads have consumed all available connections, the generator will hang indefinitely while waiting for a connection
• Since the ID generator is hung, all of the connection threads hang too leading to a deadlock

The relevant portion of the stack trace seems to be:

```
at sun.misc.Unsafe.park(boolean, long)
at java.util.concurrent.locks.LockSupport.park(java.lang.Object) (line: 175)
at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await() (line: 2039)
at org.apache.commons.pool2.impl.LinkedBlockingDeque.takeFirst() (line: 583)
at org.apache.commons.pool2.impl.GenericObjectPool.borrowObject(long) (line: 442)
at org.apache.commons.pool2.impl.GenericObjectPool.borrowObject() (line: 363)
at org.apache.commons.dbcp2.PoolingDataSource.getConnection() (line: 134)
at org.apache.commons.dbcp2.BasicDataSource.getConnection() (line: 1533)
at org.hibernate.engine.jdbc.connections.internal.DatasourceConnectionProviderImpl.getConnection() (line: 122)
at org.hibernate.internal.NonContextualJdbcConnectionAccess.obtainConnection() (line: 35)
at org.hibernate.resource.transaction.backend.jdbc.internal.JdbcIsolationDelegate.delegateWork(org.hibernate.jdbc.WorkExecutorVisitable, boolean) (line: 48)
at org.hibernate.id.enhanced.TableStructure$1.getNextValue() (line: 125)
at org.hibernate.id.enhanced.PooledOptimizer.generate(org.hibernate.id.enhanced.AccessCallback) (line: 73)
at org.hibernate.id.enhanced.SequenceStyleGenerator.generate(org.hibernate.engine.spi.SharedSessionContractImplementor, java.lang.Object) (line: 432)
at org.hibernate.event.internal.AbstractSaveEventListener.saveWithGeneratedId(java.lang.Object, java.lang.String, java.lang.Object, org.hibernate.event.spi.EventSource, boolean) (line: 105)
at org.hibernate.jpa.event.internal.core.JpaPersistEventListener.saveWithGeneratedId(java.lang.Object, java.lang.String, java.lang.Object, org.hibernate.event.spi.EventSource, boolean) (line: 67)
```

I guess the question on my mind is why does this operation need to proceed in a separate connection? That just seems ripe for this kind of deadlock.