Recovery performance is normally be influenced by
parallel_execution_message_sze. Bump this up to 32k or 64k. Also look at
v$recovery_progress
Cheers,
Leng
On 26 Sep 2019, at 4:21 am, Rakesh Ra <rakeshra.tr@xxxxxxxxx> wrote:
Hi All,
The active DG 11.1.0.7 database is running slow when there is loads happening
in the primary. The DBWR processes are waiting for "latch free" wait event.
Using Tanel Poder's latchprofx script has helped to identify that the below
latches are being acquired
NAME Held Gets Held %
Held ms Avg hold ms
--------------------------------------------- ---------- ---------- -------
----------- -----------
active checkpoint queue latch 6965 6961 6.97
14951.069 2.148
messages 674 673 .67
1446.808 2.150
checkpoint queue latch 95 95 .10
203.927 2.147
cache buffers chains 57 57 .06
122.356 2.147
Is there anyone who has experienced similar issue and had a workaround to fix
the same? I apologize that the version of DB is very old, but if you do
recollect something related to similar issue, it would be helpful if you can
shed some light here.
I did observe one note in MOS "
Bug 12314102 - checkpoint queue latch contention during media recovery (Doc
ID 12314102.8)" but don't have enough data to prove the same.