- Joined
- 10 February 2000
- Messages
- 2,677
Sorry for the down-time today. An automated update to the database server apparently went badly in the middle of the night. It created a real mess and corrupted all the databases, including the database server's own operational database, and thus the db server wouldn't even start up again.
While restoring the site to the last backup would have brought it back up pretty quickly, I really didn't want to revert to that because it would have resulted in the loss of data since the backup was created.
After going down a bunch of dead ends trying to figure out what was wrong, and then how to fix it, in the end it was recovered by creating a new "sandbox" database server, importing the broken databases, repairing them, exporting them, saving all the production database server config, doing a clean database server installation on the production server, and re-importing the fixed databases and configuration. All these steps (identifying the problem, figuring out how to fix it, and then the actual process of fixing it) took a considerable amount of time, but I felt that was preferable to losing the most recent data.
While restoring the site to the last backup would have brought it back up pretty quickly, I really didn't want to revert to that because it would have resulted in the loss of data since the backup was created.
After going down a bunch of dead ends trying to figure out what was wrong, and then how to fix it, in the end it was recovered by creating a new "sandbox" database server, importing the broken databases, repairing them, exporting them, saving all the production database server config, doing a clean database server installation on the production server, and re-importing the fixed databases and configuration. All these steps (identifying the problem, figuring out how to fix it, and then the actual process of fixing it) took a considerable amount of time, but I felt that was preferable to losing the most recent data.