Printable View
At busy insert time we have log switches (current redo log=250 MB) every 2-3 mins. To increase the log switches interval we were considering increasing the redo log size to 500MB or even 1 Gig.
Does anyone know of any dis-advantages to that?
Does anyone have redo logs this big?
Any other suggestions to increase redo log switch interval?
That's a pretty big redo log already. You have to weigh the time you are doing a log switch with the amount of data that you will potentially lose if the log file goes away...
2-3 minutes for a peak-hours doesn't seem to frequent to me. What's the log-switch interval during normal load? Do you have any complaints about "checkpoint not complete" or "archiving not be fast enough" in your alert log?
If you realy want to set a redo log so big (500M or 1G) I'd advise you to perform a couple of explicit checkpoints during filling of your online log (set LOG_CHECKPOINT_INTERVAL and/or LOG_CHECKPOINT_TIMEOUT properly). If you do not do so you'll potentially face two problems:
1. In case you loose all members of your active logfile group you'll loose quite a big amount of data (1G worth of transactions).
2. In case your database crashes or your performs shutdown abort your instance recovery after restart might take quite a long time.