Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: LOBs, PCTVERSION, and ORA-22924: snapshot too old
I've had a little play with this, and the problem seems
to be standard read-consistency.
I've created a table with a LOB column (trying both in-line enabled and disabled), and inserted one row.
This generates 3 segments:
The table segment
The LOB segment
The LOB_INDEX segment
The LOB segment blocks appear to be controlled by a different read-consistency model from the other blocks in that there is no ITL section (which is reasonable since any one block should be allocated to one version of one LOB, and is not shared with multiple other LOBs). The LOB segment simply has a commit number on it.
The LOB_INDEX segment, however, has the normal ITL entries on it, and is therefore presumably subject to exactly the same read-consistency mechanisms of the rest of the database.
The Test:
Session 2 - whilst session 1 is sleeping, start inserting, deleting and commiting lots of rows to the database to run round all the available rollback, and guarantee that every block has been overwritten (total rollback blocks 128, total rollback blocks written ca. 1,600)
Result:
Follow-up
Conclusion:
--
Jonathan Lewis
Yet another Oracle-related web site: www.jlcomp.demon.co.uk
Received on Wed Mar 31 1999 - 04:35:35 CST