Lab briefing · Stop it, start it, break it

Take an instance through EVERY state, then engineer THREE STARTUP FAILURES and diagnose each FROM THE ALERT LOG ALONE. "From the alert log alone" is the constraint that makes this worth doing. You will have engineered the failure, so you know the answer — and the exercise is to reach it the way you would at 3am on a system you did not break. FIRST, WALK THE STATES DELIBERATELY. NOMOUNT, then MOUNT, then OPEN, stopping at each and asking what is possible here. Query something in MOUNT and watch it refuse; look at the control file's view of the datafiles in MOUNT and notice you can see what the database has without having opened it. That is the state recovery happens in, and doing it once by hand is what makes Module 5 read as a procedure rather than as an incantation. THEN SHUT DOWN THREE WAYS AND TIME EACH.…

In this module: Module 3 · The database itself

1. 1Instance and database: two different things
2. 2Starting and stopping, properly
3. 3Tablespaces and datafiles
4. 4Redo: the reason recovery is possible
5. 5Archive log mode, and the day it was off
6. 6Undo, and read consistency
7. 7Parameters, and the ones that matter here
8. 8The data dictionary, as a diagnostic tool
9. 9Reading an error you have never seenFree preview
10. 10What breaks: four database-tier failures
11. 11Lab briefing · Stop it, start it, break it