Another Recovery from Hell story

Having read Howard Rogers's story and blogged about it earlier, I also came across Connor McDonald's Hell Story.

SQL Statement Execution Times

Daniel Fink has posted a blog entry about how Execution Statistics for long running SQLs might not be visible in StatsPack reports if the SQLs end after the StatsPack Snapshots. The general advice is to NOT run StatsPack Snapshots at long intervals ("intervals of 8 hours are meaningless" we frequently hear). However, if there was a long running SQL that began before the first StatsPack Snapshot and ended after the last Snapshot, that SQL wouldn't be seen in the StatsPack Report at all. You would be left wondering "if the server was so busy [and 'sar' did show 'busy'ness] why don't I see any significant SQLs in the Report". You might well miss the SQLs accounting for most of the buffer gets, disk reads, CPU time.

"Recovery" in Cold Backup

In a recent email discussion thread on ORACLE-L, I outlined methods
of "recovery" from loss of control files(s) and online redo logs IF
the last shutdown had been a Normal Shutdown.
(eg if ControlFiles and/or RedoLog files have been deleted after the
shutdown --- such a strategy could also be used where the database has
been cloned by copying only the datafiles from cold backup (shutdown normal/immediate).