Hibernate version: 2.1.6

Hey guys, I'm running into a strange problem with multiple threads in my Hibernate webapp. Each HttpServletRequest gets its own Session from the shared SessionFactory. When only one thread is active at a time, everything is fine. When two threads are active at the same time, one of them almost inevitably throws a LazyInitializationException while traversing the object graph, saying "no session or session was closed".

I initially thought that perhaps both Sessions were using the same underlying Connection object. However, I've tried no connection pooling, c3p0, proxool, and manually getting a Connection object from DriverManager and calling sessionFactory.openSession(conn), and get the same results every time.

I'm using Oracle 9i and the thin Oracle driver. Has anyone ever seen anything like this? Can anyone suggest avenues of research?

Thanks very much for any help.

I have a situation same as yours and it occurs occasionally.I ever tried to track the trace,but got nothing about it.When i remove the attribute
lazy="true" in class tag,is works fine.First i'm sure that i have alreay taken the values of PO into a FormBean (because i use struts) and closed the session before View layer is coming,So i always don's know why it is so far.

Hi, amjn. Like you, I'm using lazy collections, but in my case they must be lazy because otherwise the entire object graph would be loaded into memory and it's way too big for that. My workaround at the moment is to put my controller into a critical section and only allow one thread in at a time, but this is obviously undesirable.

I did note something interesting after I turned on DEBUG logs and poked through them. In a few places, it seemed that hibernate was giving statistics on the number of open prepared statements and resultsets. At least once, the prepared statement count appeared to be -1. I unfortunately cannot produce the exact log line at this time since I don't have Oracle handy at home, but can look into it tomorrow if y'all think that merits attention.

I've been working on this problem off and on for a week or two now and still haven't resolved the issue. Indeed, it keeps looking more and more like a bug in hibernate itself. Can someone suggest an avenue of research that might help verify or invalidate this hypothesis? Something stupid I might be doing that I have overlooked?

you're right, there must be a bug in hibernate.
In fact you're the first user using hibernate in a multithreaded app... :)
sorry but the bug is of course in your code.

how do you open close a session? servlet filter? show code

Naturally, I'm not the first to use hibernate in a multi-threaded app! That does not necessarily mean that hibernate is entirely bug-free in this regard.

Of course, as I'm in the act of posting my code, I see the glaring flaw that had eluded me until now. I thank anthony for his kind kick in the pants, which is exactly what I needed, and will now get some real work done.

Still curious about the prepared statement count apparantly being negative at one point, but I'll dig into that more deeply if it recurs. Cheers.

friend,
there are many many webapps based on hibernate in production so i can say that hibernate is bug free about this.

It's really easy: as long as there is no code to see (and the initial posting form is not filled out completely), it can never be a bug in Hibernate. We are getting way too many bogus reports these days and it takes too much time. We got that down a little bit now, but it could still be better.

Might it be helpful if thread creators could close their threads when, as now, they become irrelevant?

I've done the open source project tech support thing before and I don't envy you guys. Overall, y'all do a damn good job.

Then you know that this thread will only continue if you provide the required information about your questions.