Oracle RAC - Remastering Questions on V$GCSPFMASTER_INFO & X$KJDRMAFNSTATS

From: PD Malik <pdthedba@xxxxxxxxx>
To: Oracle-L Group <oracle-l@xxxxxxxxxxxxx>, rshamsud@xxxxxxxxxxxxxxxx
Date: Thu, 28 Jul 2011 15:16:09 +0100

Hello experts,



Am I right in assuming that the following 2 queries should return the same
result as both are presumably telling the number of objects owned by the
current master? :

SQL> select count(*) from V$GCSPFMASTER_INFO where current_master=0;

COUNT(*)

----------

217

SQL> select my_objects from X$KJDRMAFNSTATS;

MY_OBJECTS

----------

52



Similarly, should both these queries not be returning the same result as
well as both are telling the total number of remastering happened? :



SQL> select sum(remaster_cnt) from V$GCSPFMASTER_INFO;

SUM(REMASTER_CNT)

-----------------

3108

SQL> select drms from X$KJDRMAFNSTATS;

DRMS

----------

11919

(This output just for reference :

SQL> select sum(remaster_cnt) from gV$GCSPFMASTER_INFO;

SUM(REMASTER_CNT)

-----------------

15544

)



Instance been up since 3rd July morning.



We've a 5 node busy 10TB 10.2.0.5 RAC on AIX 6L with around 15-16K sessions
running on an average day :



SQL>select count(*) from gv$session;

  COUNT(*)
----------
     15732



Also, just by looking at the numbers above can something be said on the
amount of our remastering activity please are they normal/too high/OK etc. I
understand this may be a bit vague question becs 'it depends' but still if
something can be said, it'll be great.



Can I please also ask, how much stress does remastering put on background
processes in particular LMS? Reason for asking is because reading Riyaj's
documents it seems during Remastering, GRD only transfers a new hash table
across the nodes which shudn't be a great deal (resources wise) to do and
there are no actual blocks transferred between instances during remastering
so wudn't have thought it would be heavy resource consuming. I am aware it
does freeze the GRD for a very brief time but that shouldn't impact the
normal users working who dont need GRD work at that particular time?



I am not seeing events like gc remaster or drm freeze etc. as our top events
in AWR but on the other hand there have been a couple of recent CPU blips on
the PROD boxes (CPU touched the floor and instances totally hung) for about
90 seconds and the top wait event for that period was 'gcs drm freeze in
enter server mode' but that may be a symptom or could that be a cause due to
excessive remastering please?



Thanks in advance for any help in this area please.

 :PD

Oracle RAC - Remastering Questions on V$GCSPFMASTER_INFO & X$KJDRMAFNSTATS

Other related posts: