in re: “Holders can be identified where the request is 0, as the request column specifies the type of wait request — for a holder this column has no meaning. Waiters, likewise, can be seen as having an lmode of 0 as that is only a valid value for a holder.”

It is possible for a holder to have a request > 0. This happens when a lock conversion is taking place where the holder of a lock is requesting a higher lock. This also means that a holder of a lower lock can also be a waiter requesting a higher lock. Your blocking sessions query will not catch this situation.

To identify waiters, I only use request > 0, then join that id1, id2 back to gv$lock to find the holder of the lock.

This is explained well in Expert Oracle Architecture – Thomas Kyte or by a search for oracle lock conversion.

Thanks, Michael Wehrle

select VIEW_DEFINITION
from V$FIXED_VIEW_DEFINITION
where view_name=’V$SESSION’;

If you don’t have any, you may want to look at that post http://www.pythian.com/blogs/968/installing-oracle-11g-on-ubuntu-804-lts-hardy-heron (I should work with 10g too)

Regarding your other question you may want to look at V$GES_ENQUEUE. And with RAC you could be interested to look at US, TA enqueues and probably many more.

The LOCK word means too many things (And not only with Oracle). You could use the V$ENQUEUE_LOCK fixed view instead. BTW, Why to you display a “KILL SESSION” command ?