Hi Frank,
For this instance, the peak rate for new connections is maybe 6/sec, but
only for 1 second. Most of the time, there are around 30 new connections
every minute when PSP0 is in a CPU wait.
These aren't huge numbers of connections to me. Connection pooling could
possibly solve most of this, but as the end users aren't affected, I see no
value to offset the implementation costs for it.
Thanks!
Rich
On Fri, Feb 12, 2021 at 3:54 AM Frank Gordon <frankagordon@xxxxxxxxx> wrote:
Hello,
Is your application spawning a "huge amount" of database server processes?
Regards,
Frank
On Wed, Feb 3, 2021 at 5:58 PM Rich J <rich242j@xxxxxxxxx> wrote:
Hey all,
In one of our 19.6 EE instances (OL7.7 VM), the PSP0 background process
occasionally goes into a CPU wait state for an extended time (>1 hour).
This shows up both in real-time monitoring with EM13.4 as well as AWR
reports as event "CPU + Wait for CPU". However, there is no correlated CPU
usage from the OS side. There are several "rdbms ipc message" idle waits
reported on this process.
There are no trace files generated, there are no entries in the alert log
and nothing about this that I could find in MOS. Other than EM/AWR saying
that PSP0 is waiting, I don't see any negative effects of this event.
Thoughts, anyone? I guess I'm wondering "If it ain't broke..." applies
or is this a precursor/omen of fun things to come. Or is this another case
of an idle wait being treated as non-idle in reporting?
Thanks,
Rich
--
+353-86-0695383
