OK, bring in the DB Techies
Posted: Mon Sep 05, 2005 1:50 pm
This is not intended to be a whine at GOA - I have 3 characters stripped of all items but luckily had their templates stored so I am confident they will be fixed in time. I just cant get my head round why GOA are fixing the problems in this way.
For information I work as a professional DBA - I have been fixing database problems for more years than I can remember.
I read somewhere that the DAOC database is not a 'relational' database (such as Oracle, DB2, SQLServer etc.) but rather a 'flat-file' database of not specified origin. However, technically, the principles for this discussion should still apply.
I'll try to explain in less technical terms.....
When we 'relational' DBA's talk about a rollback, what we really mean is a restore and rollforward to a point in time prior to a problem occurring hence it appears as a rollback before the problem occurred.
Consider the scenario:
A problem occurred on a database at time 'X' which could not be fixed without implementing a 'rollback' (this could be hardware failure, software related, human error or any numerous reason).
What happens next is a full backup of the database (usually backups are taken daily/weekly and stored on a tape cartridge away from the actual database server) is loaded over the top of the faulty database thereby replacing it with an un-faulty albeit older version.
Then transactional logs which store all database transactions (these logging activities are occurring when you see the 'your character has been saved' message in game) between successful backups are re-applied to the older restored database until a point in time just before the problem occurred.
Hence if GOA have performed such a rollback, the only problems we should be encountering with our charactares is the loss of items/exp/money obtained in the period of time between the point in time of the rollback and the time the database was shutdown to perfoirm the rollback which according to GOA should be about half a day (Wednesday). I'm sure we would all prefer this scenario rather than thousands of RightNow submissions and trying to remember what is on our characters and vaults.
I would pose some questions with the current GOA solution:
(1) Did they have a problem with their tape backups? maybe they didn't have a valid backup that they could restore?
(2) Did the problem corrupting the database happen a long time ago and only an event more recent has caused the database to actually 'show' the corruptions? ie. they could not do a rollback because the period of data-loss would be unacceptable to all customers.
(3) Has a human-error occurred? eg. Haxxor or accidental GOA employee related?
All the above cases would need a manual data-cleanse which seems to be what is happening here and in business terms a DBA somewhere would be looking for a new job because of this.
For information I work as a professional DBA - I have been fixing database problems for more years than I can remember.
I read somewhere that the DAOC database is not a 'relational' database (such as Oracle, DB2, SQLServer etc.) but rather a 'flat-file' database of not specified origin. However, technically, the principles for this discussion should still apply.
I'll try to explain in less technical terms.....
When we 'relational' DBA's talk about a rollback, what we really mean is a restore and rollforward to a point in time prior to a problem occurring hence it appears as a rollback before the problem occurred.
Consider the scenario:
A problem occurred on a database at time 'X' which could not be fixed without implementing a 'rollback' (this could be hardware failure, software related, human error or any numerous reason).
What happens next is a full backup of the database (usually backups are taken daily/weekly and stored on a tape cartridge away from the actual database server) is loaded over the top of the faulty database thereby replacing it with an un-faulty albeit older version.
Then transactional logs which store all database transactions (these logging activities are occurring when you see the 'your character has been saved' message in game) between successful backups are re-applied to the older restored database until a point in time just before the problem occurred.
Hence if GOA have performed such a rollback, the only problems we should be encountering with our charactares is the loss of items/exp/money obtained in the period of time between the point in time of the rollback and the time the database was shutdown to perfoirm the rollback which according to GOA should be about half a day (Wednesday). I'm sure we would all prefer this scenario rather than thousands of RightNow submissions and trying to remember what is on our characters and vaults.
I would pose some questions with the current GOA solution:
(1) Did they have a problem with their tape backups? maybe they didn't have a valid backup that they could restore?
(2) Did the problem corrupting the database happen a long time ago and only an event more recent has caused the database to actually 'show' the corruptions? ie. they could not do a rollback because the period of data-loss would be unacceptable to all customers.
(3) Has a human-error occurred? eg. Haxxor or accidental GOA employee related?
All the above cases would need a manual data-cleanse which seems to be what is happening here and in business terms a DBA somewhere would be looking for a new job because of this.