Hi,
Iam relatively new to hibernate. Encountering a weird scenario where multiple threads are working on the same parent object. The parent has a <one-to-many> child Map (inverse="true"). Basically each of the thread adds different values of the child attribute with the same key. For eg.
ParentA initially has no child attributes.
Thread1 -> ParentA -> (key1, value1)
Thread2 -> ParentA -> (key1, value2)

The logic determines if there are any values for the key, and if it's null creates a new child object with the key/value and sets the parent instance. If value is a non-null it updates the value.
So in the above scenario there are two child objects created with the same key. When Thread1 updates the parent, the child objects is also persisted. When Thread2 updates there is a StaleObject thrown, there is retry logic that refreshes the session, and reloads the data. So there should be ONLY one child object key1 with value2. But there are two records key1, value1 and key2, value2 persisted in the database.
Any suggestions?

Can you please post the code of this your retry logic.
Do you also use Cascade on persist option?

Apologize for not posting it earlier. Didn't want to spam with code.
The gist of the retry logic is as follows-
boolean success = false;
while (!success) {
try {
processOlapQuery(idIterator, outputFileHandler, helper, operationHandler);
success = true;
} catch (StaleObjectStateException e) {
//first we'll restart session
DAOSessionHelper.endSession();
DAOSessionHelper.startSession();
//then reset input, output, and helper/handler
idIterator.reset();
outputFileHandler.reset();
helper.reset();
operationHandler.refresh();
//check this AFTER we reset everything, so session is in a usable state
if (staleStateCount > MAX_STALE_STATE_RETRIES) {
logger.error("Un-recoverable error (too many stalestates!): " + e.getMessage(), e);
throw e;
}
staleStateCount++;
}
}

And yes the map property of the child in the parent config file has the cascade as
cascade="all,delete-orphan"