On 12/07/2017 04:18 PM, Mark W. Farnham wrote:
By far the most frequent cause of the problems with LGWR are too small log files. I have once seen a customer which had 3 checkpoints per minute. Of course, redo log files were 50M in size. Furthermore, duplexing redo logs is completely unnecessary. Duplexing is done by the storage (RAID 1+0 or some other form of RAID) and it is unnecessary to waste the expensive server CPU and IO capacity to do the job that storage does anyway.
There are a few avenues of research to discover why LGWR might be identified as blocking.