Re: initjvm.sql hangs on db file sequential read to IDL_UB1$

  • From: Jeremiah Wilton <jeremiah@xxxxxxxxxxx>
  • To: Adam Donahue <adonahue@xxxxxxxxxxx>
  • Date: Tue, 4 Jan 2005 16:31:17 -0800 (PST)

Adam wrote:

> I've having trouble installing the Oracle JVM on an Oracle
> instance on Redhat Linux AS 3.0.  I thought perhaps some of you would
> have ideas.
> ...
> I then execute @?/javavm/install/initjvm
> which runs fine and gets to the CREATE JAVA SYSTEM statement.  A trace
> reveals this is running fine.  At some point, however, it begins waiting on:
>    db file sequential read
> against the
>    IDL_UB1$
> table.  (Found via p1, p2 in v$session_wait.)  Actually, not only the
> same table, but the same /block/.  It doesn't report any errors, nothing
> in the alert log, it just hangs.


I can't address Oracle JVM specifically, but I can speak to the type
of hanging problem you are encountering.

When you are querying v$session_wait, you may notice that sequence# is
not incrementing.  This means that the session is not really waiting
on that event, only that that was the last event posted before the
session hung.

You need to determine first if the process is spinning or hung.  Look
at top or ps for the server process of your session, and see if it is
using any CPU or not.

If yes, then it is spinning.  If no then it is hung.  Most likely in
your case it is hung.

You can obtain an errorstack of the hung process using oradebug:

SQL> oradebug setospid <pid of server process>
SQL> oradebug dump errorstack 4

Then you can paste the call stack into the metalink call stack matcher
to see if it is a known bug:

Or submit the errorstack to support.

If it is spinning, then support will want several errorstacks taken at
short intervals.

You may also have some luck using strace on the server process to see
what it is stuck or spinning on.

Good luck,
Jeremiah Wilton
ORA-600 Consulting
Emergencies - Seminars - Hiring
On Tue, 4 Jan 2005, Adam Donahue wrote:


Other related posts: