We are running Oracle 8.1.6 (64-bit) on solaris. Our redo logs are "raw" devices
sized at 60Meg. We are experimenting with setting the LOG_CHECKPOINT_INTERVAL
and are encountering some "weird" behavior. Even setting the value as low as 100
doesn't seem to cause checkpoints (Checkpoints are being logged to alert) and we
are getting checkpoints at only log switches. Is this a bug or are we doing something
wrong here? I thought the block size was 512 for solaris and hence this should have
casued a checkpoint at every 51200 bytes of redo written out? Is this nay different
for "raw" files? Please respond! Thanks!
You do have LOG_CHECKPOINTS_TO_ALERT = TRUE, right?
Yes, the block size on solaris is typically 512 bytes. It looks as though you have the LOG_CHECKPOINT_INTERVAL set correctly if you want it to checkpoint that often. Is LOG_CHECKPOINT_TIMEOUT set to something funky maybe?
Your Log_checkpoint_intreval should be 122880 instead of 512000.It should be Logfilesize/osblocksize.
2) You can also make log_timeout_intreval=0
If you set the above said values, your log switch should happened based on Log file size. If you are running heay OLTP intensive application, you can encounter 4 log switches per hour. This is good as for as performance issues are concerned.
Sundar,Oracle DBA, Telesis