Hello Manfred,
at first you may need to understand what "private strand flush" stands for. You
can check out one of my old blog post about this:
http://tinyurl.com/j6p9uu6
You should check your transaction with used private redo size in x$ktifp (as it
also can fall back to the old behavior if redo gets too large or you
use different features like flashback database, etc.) or why the DBWR wants to
write these associated blocks so fast (e.g. incremental checkpoints or
running out of usable buffer cache blocks, etc.), if you have a lot of "Private
strand flush not complete" messages.
Secondly you may should think about increasing the redo log group size, if you
see a switch every minute (or more), but this is something different as
well.
Best Regards
Stefan Koehler
Freelance Oracle performance consultant and researcher
Homepage: http://www.soocs.de
Twitter: @OracleSK
Pausch Manfred <Manfred.Pausch@xxxxxxxxxxxxx> hat am 7. April 2016 um 10:57--
geschrieben:
Hello all,
I consider myself as an „advanced beginner“ to Oracle database
administration so you may excuse if
my question should be absurd ;-)
In one of our dev databases (Oracle 12c on Linux) we frequently (about once
a minute) see
Thu Apr 07 10:42:24 2016
Thread 1 cannot allocate new log, sequence 71029
Private strand flush not complete
Current log# 2 seq# 71028 mem# 0:
+ORAREDO1/DEVEE072/ONLINELOG/group_2.269.854394251
Current log# 2 seq# 71028 mem# 1:
+ORAREDO2/DEVEE072/ONLINELOG/group_2.269.854394251
Beginning log switch checkpoint up to RBA [0x11575.2.10], SCN: 4102856777
Thu Apr 07 10:42:27 2016
Thread 1 advanced to log sequence 71029 (LGWR switch)
Current log# 3 seq# 71029 mem# 0:
+ORAREDO1/DEVEE072/ONLINELOG/group_3.268.854394253
Current log# 3 seq# 71029 mem# 1:
+ORAREDO2/DEVEE072/ONLINELOG/group_3.268.854394253
We also had “cannot allocate new log” together with “checkpoint not
complete”.
But seems that we got rid of this by increasing db_writer_processes from 2
to 4.
(we did the change yesterday, so we still observing this)
According to MOS note 372557.1 :
“These [private strand flush not complete] messages are not a cause for
concern unless there is a significant time gap between
the "cannot allocate new log" message and the "advanced to log sequence"
message.”
In our case the time gap between this two messages regularly is 3seconds.
Would anybody of you consider this as “significant time gap”?
Many thanks
Manfred
