Just a quick question really. I should know the answer but having a mental block!
IF OE Replication is enabled, and we've got an AI Archive interval of, say, 15 minutes, if REPL goes down does that mean we have 15 minutes per AI extent to get Replication back up and running because otherwise all the AI extents will be locked pending transfer to target? Then the DB will stop unable to switch AI extents?
Yes. If you are using variable length extents then the last one will continue to grow and the DB will not stop until the AI extents fills its file system. WIth fixed length extents, your DB will stall (with -aistall) or crash if it fills the last extent and cannot switch.
Yes. If you are using variable length extents then the last one will continue to grow and the DB will not stop until the AI extents fills its file system. WIth fixed length extents, your DB will stall (with -aistall) or crash if it fills the last extent and cannot switch.
Thanks Paul - good to know the last one will continue to grow if it's variable.
So would you set the archive interval higher for a replication enabled database as the REPL target is likely to be the primary DR solution?
You can always switch the archive interval to 'on demand' (0) or a longer interval (value in seconds) when you know you're in this situation:
rfutil james -C aiarchiver setinterval 0
True, it's more for when you don't know you're in the situation though! :)
Or if you switch every 15 mins, you might add sufficient amount of AI files to get you thru at least 24 hrs in case such a situation happens, ie 96 in this case, so you have a good buffer, and still would be able to make a backup for example while the 'replication down' is being fixed.
DR is one reason for AIs but the other reason I recall is to be able to recover from bad data - ie someone goes in and empties a table accidentally. That accident would replicate just fine. But you use your AIs to restore until point in time. So do you consider that before increasing your interval?
The interval would have little to do with recovering from bad data. You'd use a roll forward time or transid to recover and this would involve rolling forward one or very many ai files in order to do so ..