Restore Aurora DB using BIn-log position

Question

A customer is considering migrating their DB to Aurora,
one of their requirements is to be able to restore the DB to an earlier state based on Bin-log position,
meaning to restore the DB to a prior state based on what function were ran last.

so basically very similar to "backtrack" but based on functions that ran and not point in time.

Accepted Answer

Ok, it's an interesting concept but not without risk, there's questions about consistency if they start cherry-picking what changes out of the binlogs to apply.

To clarify Aurora capabilities:

1. you cannot access the actual binlog files
 2. you can attach as a replication slave and request binlog records from a given position. The slave can be another Aurora or MySQL instance/cluster replaying the logs, DMS, binlog proxy (eg. MaxScale) or some other type of binlog parser
 3. Aurora does not support filtered replication

Since we're talking about DR strategy, they should ship the binlogs off the Aurora cluster anyway for this particular use case, so #2 above comes into play. Once they have them externally to the Aurora cluster on a proxy or something they can process/analyze them as needed, but it would be a manual process.

They would need to test it out. I don't know of any customers that do that today, so it's hard to say whether they will encounter other gotcha's or not.

Restore Aurora DB using BIn-log position

Contenido relevante