Data-guard conceptual help needed

Question

Hello Experts,
We have 2-node RAC Physical standby DG. I got the following doubts in my mind just to clear my concept:-
1. Can we perform Hang analyze on the DG site, when it is on managed recovery and of-course in read-only mode? if not then Q2?
2. If both of my DG nodes (in a 2-node RAC DG) are in a hung state while the managed recovery was active, can I do sqlplus sql> set _prelim on and enable oradebug hanganalyze or do I have any other option apart from restarting the DB
3. IF the Primary RAC (2-node) hangs then if we take systemstate dump / Hanganalyze from the DG site
I guess, we will not get meaningful results as Hanganalyze will use Kernel system calls of the Server on the DG site and systemstatsdump will dump the SGA of the secondary site instance's SGA
4. I am using a non-RAC primary and Secondary site and non-ASM disk
I have my Primary mount -points as /Oradata/Prim1 = 1 TB and Physical standby DG - real time apply mount-point as /Oradata/Seco1 =.8 TB & /Oradata/Seco2 = .2TB and my Primary DB is now on size expansion from .79 TB to .95 TB overnight. I am not intending not to take any down-time nor am I able to extend the disks on the Primary for any reason. what can be done wisely to keep the DR site in close sync?
Regards,
Syvalli
Edited by: user4566776 on Dec 12, 2012 6:04 AM

Answer 1

We have 2-node RAC Physical standby DG. I got the following doubts in my mind just to clear my concept:-
1. Can we perform Hang analyze on the DG site, when it is on managed recovery and of-course in read-only mode? if not then Q2?
2. If both of my DG nodes (in a 2-node RAC DG) are in a hung state while the managed recovery was active, can I do sqlplus sql> set _prelim on and enable oradebug hanganalyze or do I have any other option apart from restarting the DBYes you can perform hang analyze and even you can generate active session history report if DR is on read-only mode.
SQL> select database_role from v$database;
DATABASE_ROLE
PHYSICAL STANDBY
SQL>
SQL> oradebug setmypid;
Statement processed.
SQL> oradebug unlimit;
Statement processed.
SQL>
SQL> oradebug hanganalyze 3;
Hang Analysis in /u02/app/oracle/diag/rdbms/CKPT_un/CKPT/trace/CKPT_ora_5113.trc
SQL>
SQL> oradebug dump systemstate 266;
Statement processed.
SQL>
[oracle@oracle-stby ~]$ tail -10 /u02/app/oracle/diag/rdbms/CKPT_un/CKPT/trace/CKPT_ora_5113.trc
  Holder:
  SO: 0x8ca2c8b8, type: 83, owner: 0x6000c478, flag: INIT/-/-/0x00 if: 0x3 c: 0x3
   proc=0x90488f50, name=circuit holder, file=kmc.h LINE:2615, pg=0
  (circuit holder) disp = 0x8ca2c830 (0, 1), proc = (0x90488f50, 1)
END DISPATCHER DUMPS
END OF SYSTEM STATE
*** 2012-12-12 20:03:44.100
Oradebug command 'dump systemstate 266' console output: <none>
[oracle@oracle-stby ~]$
3. IF the Primary RAC (2-node) hangs then if we take systemstate dump / Hanganalyze from the DG site
I guess, we will not get meaningful results as Hanganalyze will use Kernel system calls of the Server on the DG site and systemstatsdump will dump the SGA of the secondary site instance's SGAYes, the hang analysis we should perform from the instance where you have problem. you can use OS level trace if database not accessible by using How to interpret OS system traces [ID 1429678.1]
4. I am using a non-RAC primary and Secondary site and non-ASM disk
I have my Primary mount -points as /Oradata/Prim1 = 1 TB and Physical standby DG - real time apply mount-point as /Oradata/Seco1 =.8 TB & /Oradata/Seco2 = .2TB and my Primary DB is now on size expansion from .79 TB to .95 TB overnight. I am not intending not to take any down-time nor am I able to extend the disks on the Primary for any reason. what can be done wisely to keep the DR site in close sync?Can you tell me one thing? After size expanded in primary from .79 TB to .95 TB does your standby is out of SYNC?
If you do not have space, then you should add some space to the disk to avoid any interruption to production. Or check whether you can adjust HWM on primary database using below method. http://www.oracle-ckpt.com/how-to-shrink-datafiles-adjusting-hwm/
Hope this helps