[nhibernate-issues] [NHJIRA] Created: (NHSR-37) Dead Lock: delete with optimize

[RAB (JIRA) <nh...@gm...>]

Dead Lock: delete with optimize
-------------------------------

                 Key: NHSR-37
                 URL: http://216.121.112.228/browse/NHSR-37
             Project: NHibernate.Search
          Issue Type: Bug
          Components: Core
    Affects Versions: 2.1.0
            Reporter: RAB
            Assignee: Paul Hatcher
            Priority: Critical


With this option, the index is optimized for each transaction.
x.SetProperty("hibernate.search.default.optimizer.transaction_limit.max", "0");

Delete workflow:
•	Delete-Work is performed.
•	Delete-Work opens an index reader and lock the provider.
•	Delete-Work is finished.

•	CleanUp is performed.
•	All index readers are closed.

•	CleanUp starts the optimizer.
•	Optimizer opens an index writer and lock the provider.

•	Back on the CleanUp method.
•	All index writer are closed.

•	For each locked provider Monitor.Exit (provider) is called.
•	Finish.

Monitor.Enter(provider) was executed twice but Monitor.Exit(provider) only once.
Delete the next time we have a dead lock.

Line 204 - 224 LockProvider in Backend\Workspace
Now:
private void LockProvider(IDirectoryProvider provider)
{
	// Make sure to use a semaphore
	object syncLock = searchFactoryImplementor.GetLockableDirectoryProviders()[provider];
	Monitor.Enter(syncLock);
	try
	{
		if (!lockedProviders.Contains(provider))
		{
			lockedProviders.Add(provider);
			dpStatistics.Add(provider, new DPStatistics());
		}

	}
	catch (Exception)
	{
		// NB This is correct here, we release the lock *only* if we have an error
		Monitor.Exit(syncLock);
		throw;
	}
}

Solving:
private void LockProvider(IDirectoryProvider provider)
{
	//Provider only lock once
	if (!lockedProviders.Contains(provider))
	{
		// Make sure to use a semaphore
		object syncLock = searchFactoryImplementor.GetLockableDirectoryProviders()[provider];
		Monitor.Enter(syncLock);
		try
		{
			lockedProviders.Add(provider);
			dpStatistics.Add(provider, new DPStatistics());
		}
		catch (Exception)
		{
			// NB This is correct here, we release the lock *only* if we have an error
			Monitor.Exit(syncLock);
			throw;
		}
	}
}

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://216.121.112.228/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira