In previous post I uploaded AWR reports for 4/18 (normal day, baseline), and 4/27 (slow day).
It can be tricky to pick the right times to compare.
If we compare wrong periods, we might blame slowness on increase in workload.
Picking 4/18 as our baseline was deliberate, by most metrics the DB did similar or less work on 4/27.
Per Load Profile on our slow day we see:
– similar or less logical and physical IO…although with 29% more “Read IO requests”
– less user calls, executions, and logons
– slightly less number of parses and a bit more hard parses
Per Time Model Statistics on slow day:
– DB time almost doubled from 5 million seconds to 9.2 million seconds
– DB time gained 2.2 million seconds from “hard parse elapsed time”, and 1.4 million from wait event “db file sequential read”
– higher DBU CPU added another 400 thousand seconds to DB time
What hit us?
Is it the bigger load on the db?
Is it slower disks from increase in read IOPS?
Is it the bump in CPU utilization from hard parsing?
Is it a bird…? Is it a plane…?