SQLiteDistributedLock doesn't play right with async

[fnajera-rac-de]

The following snippet shows a (valid?) usage of SQLiteDistributedLock which leaves/can leave the lock alive forever:

public async Task Method(PerformContext context)
{
    using (context.Connection.AcquireDistributedLock(..., TimeSpan.FromMinutes(1)))
    {
        // await calls here
    }
}

Expected: since the lock is in a using block, it should be released when the using loses scope.

Actual: the lock is held indefinitely.

I believe the problem resides in the use of ThreadLocal in SQLiteDistributedLock.

https://github.com/raisedapp/Hangfire.Storage.SQLite/blob/251983b6c194b0098393e60b4936fd16d0e1671f/src/main/Hangfire.Storage.SQLite/SQLiteDistributedLock.cs#L17-L18

When using async, there's no guarantee that you resume in the same thread you were before - even less in ASP.NET Core where there's no sync context.

I'll try to add a test case and a fix for this issue.

[kirides]

If I see that correctly, the distributed locking mechanism isn't event correct for single node servers

It should also block the current thread from re-entering a lock (no re-entrency) but it doesn't.

By making use of database commands the whole code could be simplified to...

• a simple Insert statement, on collision by key retry until timeout
• heartbeat: update where key and owner is ourself
• remove: remove where key and owner is ourself

That way the collision checks and updates work atomically on the database level.

I could code something up in the upcoming days

[fnajera-rac-de]

@kirides I did compare the sourcecode of other database plugins, they look to follow what you wrote, so that's probably the way to go. Unfortunately I don't currently have time to work on this.