Обсуждение: executeQuery() throws "Statement has been closed"

Hi everyone,


sometimes my application (jdbc-1101, hibernate, desktop application)
reports the following "Statement has been closed" error:


Caused by: org.postgresql.util.PSQLException: Die Anweisung wurde
geschlossen.
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.checkClosed(AbstractJdbc2Statement.java:2631)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.getMaxRows(AbstractJdbc2Statement.java:659)
    at
org.postgresql.jdbc4.Jdbc4Statement.createResultSet(Jdbc4Statement.java:34)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement$StatementResultHandler.handleResultRows(AbstractJdbc2Statement.java:219)
    at
org.postgresql.core.v3.QueryExecutorImpl.processResults(QueryExecutorImpl.java:1839)
    at
org.postgresql.core.v3.QueryExecutorImpl.execute(QueryExecutorImpl.java:255)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.execute(AbstractJdbc2Statement.java:559)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.executeWithFlags(AbstractJdbc2Statement.java:417)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.executeQuery(AbstractJdbc2Statement.java:302)
    at
org.hibernate.engine.jdbc.internal.ResultSetReturnImpl.extract(ResultSetReturnImpl.java:80)
    ... 19 more



When looking at the code of getMaxRows()
(https://github.com/pgjdbc/pgjdbc/blob/master/org/postgresql/jdbc2/AbstractJdbc2Statement.java
line 665):

     public int getMaxRows() throws SQLException
     {
         checkClosed();
         return maxrows;
     }


... I'm wondering why there is even checkClosed() called before
returning the value of a field?
Could that check be safely removed?

It may be the case that there are concurrent calls from multiple threads
to the same Connection object, but shouldn't that be thread-safe?

(Un)fortunately the exception is logged very seldomly, so there has not
yet a pattern emerged when they happen.


Thanks for your comments,
best regards

    -hannes

Hi Dave,


 > Well if another thread has closed the connection which I suspect is
> happening here, then the results of checkClosed is correct.

The JDBC connection is only closed when the application exits. Since the
message is "statement closed" (as opposed to "connection closed"), I
quite sure thats not happening here.

It may well be a concurrency issue, since we are sharing one JDBC
connections among multiple Hibernate-Sessions.


thanks again,

    -hannes

Hannes Erven <hannes@erven.at> wrote:

> sometimes my application (jdbc-1101, hibernate, desktop application)
> reports the following "Statement has been closed" error:

> ... I'm wondering why there is even checkClosed() called before
> returning the value of a field?
> Could that check be safely removed?

Not according to the API documentation:

http://docs.oracle.com/javase/7/docs/api/java/sql/Statement.html#getMaxRows%28%29

| Throws:
|     SQLException - if a database access error occurs or this method is called on a closed Statement

Besides, if a Statement has been closed, why would you expect it to
still be usable?

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

Dave,


 > Sharing one connection across multiple hibernate sessions seems
 > fraught with danger ?

sorry, I should have explained better. This is a read-only ("SELECT")
workload, so there are no commits etc. interfering between threads. This
is a session-per-view pattern so the indiviual sessions have minimal
size and may be closed and garbage collected independently.
Access to the backend is again pooled through a pgbouncer statement pool
(which might be another place to look at).


According to http://jdbc.postgresql.org/documentation/81/thread.html ,
JDBC connections are threadsafe and can be used by multiple threads
concurrently.

That page also reads,
"If a thread attempts to use the connection while another one is using
it, it will wait until the other thread has finished its current
operation. If the operation is a regular SQL statement, then the
operation consists of sending the statement and retrieving any ResultSet
(in full)."


As I understand it, even if one thread closes the connection, a call to
executeQuery() should either fail early or complete. It appears to me
that is is not the case in these rare instances where the exception is
thrown only when the results have already been processed successfully.


As Kevin pointed out (thanks!), getMaxRows() is required by the spec to
check whether the connection is still open.
So let me rephrase the question: why is
AbstractJdbc4Statement.createResultSet() calling getMaxResult() and
friends [which in turn check for openness] instead of directly using the
already prepared field values?


In any case, I'll see if I can come up with a test case to demonstrate.


Thanks again,
best regards

    -hannes

Hannes Erven <hannes@erven.at> wrote:

> As Kevin pointed out (thanks!), getMaxRows() is required by the
> spec to check whether the connection is still open.

It's more strict than that -- it is required to throw an exception
if the *Statement* has been closed.  Of course, closing a
Connection will automatically close all Statement objects
associated with that connection, but the Statement object can, and
often is, closed before the Connection.

So, something is closing the Statement (directly or indirectly) and
then there is an attempt to use it.  You need to figure out how
that happens.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

Hi,


 > http://jdbc.postgresql.org/documentation/81/thread.html
 >
> "If a thread attempts to use the connection while another one is using
> it, it will wait until the other thread has finished its current
> operation. If the operation is a regular SQL statement, then the
> operation consists of sending the statement and retrieving any ResultSet
> (in full)."


Consider this test case: one thread calls executeQuery("SELECT ..."),
and while that query is still executing, a second thread calls close()
on the statement.

Given the documentation above, I would expect the regular SQL statement
to complete, the statement then to be closed, and no exception thrown.


I'll attach a sample code for this scenario. With the current git
driver, this fails at getMaxRows().

When you change Jdbc4Statement.createResultset() so it uses this.maxrows
instead of getMaxRows() and hence skip checkClosed, then that test will
pass. Patch also attached, although I have no idea whether this really
is the right thing to do.


Still, I will also check my code. I think this happens when a user
closes a view while the data inside is (re)loading, so I'll probably
have the error handling discard that exception in this case instead of
ringing the alarm bell...


Thanks for your comments, David and Kevin!

Best regards,

    -hannes