Wednesday, April 15, 2015

How to monitor restore/recovery operation..?

Tips and techniques for monitoring a restore/recovery operation and determine whether it is indeed working, slow or hung.

In general, a restore should take approximately the same time as a backup, if not longer. Therefore,
if your backup took 10 hours to complete, it will take at least 10 hours to restore to the same host.
Another good indicator is to determine the duration of your previous restore/recovery operations.
Monitor the logs and views and observe the rate of change. Restore and recovery operations are quite resource intensive so it's important to understand if the process is working or hung.





RMAN Log

By default the result of an RMAN operation is written to the standard output. There is no default log file.
You will need to capture the result using either the SPOOL LOG option, or re-direct the standard output to a file.

Example of a restore datafile session:


RMAN> spool log to res5.out
RMAN> restore datafile 5;
RMAN> spool log off
$ cat res5.out

RMAN> restore datafile 5;

Starting restore at 27 DEC 2011 14:05:03 [1]
using target database control file instead of recovery catalog
allocated channel: ORA_DISK_1
channel ORA_DISK_1: sid=143 devtype=DISK [2]

channel ORA_DISK_1: starting datafile backupset restore
channel ORA_DISK_1: specifying datafile(s) to restore from backup set
restoring datafile 00005 to /opt/app/oracle/oradata/ORA102/example01.dbf [3]
channel ORA_DISK_1: reading from backup piece /opt/app/oracle/fra/ORA102/backupset/2011_12_27/o1_mf_nnndf_TAG20111227T122122_7hl7dm2n_.bkp [4]
channel ORA_DISK_1: restored backup piece 1
piece handle=/opt/app/oracle/fra/ORA102/backupset/2011_12_27/o1_mf_nnndf_TAG20111227T122122_7hl7dm2n_.bkp tag=TAG20111227T122122
channel ORA_DISK_1: restore complete, elapsed time: 00:00:15 [5]
Finished restore at 27 DEC 2011 14:05:19 [6]



From the above we can see:
[1] date and time when the restore started
[2] the database's session ID in v$session - 143
[3] the datafile number and name where file will be restored to
[4] the backuppiece name and TAG
[5] the time taken for the restore of this datafile, and the channel used
[6] date and time when restore completed


Example of a recovery session:


RMAN> recover datafile 5;

Starting recover at 27 DEC 2011 14:05:55
using channel ORA_DISK_1

starting media recovery

channel ORA_DISK_1: starting archive log restore to default destination [1]
channel ORA_DISK_1: restoring archive log [2]
archive log thread=1 sequence=77
...
channel ORA_DISK_1: restoring archive log
archive log thread=1 sequence=89
channel ORA_DISK_1: reading from backup piece /opt/app/oracle/fra/ORA102/backupset/2011_12_27/o1_mf_annnn_TAG20111227T135926_7hlf4jrk_.bkp
channel ORA_DISK_1: restored backup piece 1
piece handle=/opt/app/oracle/fra/ORA102/backupset/2011_12_27/o1_mf_annnn_TAG20111227T135926_7hlf4jrk_.bkp tag=TAG20111227T135926
channel ORA_DISK_1: restore complete, elapsed time: 00:01:35
...
channel default: deleting archive log(s)
archive log filename=/opt/app/oracle/fra/ORA102/archivelog/2011_12_27/o1_mf_1_88_7hlfmkr4_.arc recid=116 stamp=770998049
channel default: deleting archive log(s) [3]
archive log filename=/opt/app/oracle/fra/ORA102/archivelog/2011_12_27/o1_mf_1_89_7hlfmbd8_.arc recid=104 stamp=770998043
media recovery complete, elapsed time: 00:00:01 [4]
Finished recover at 27 DEC 2011 14:07:32 [5]

[1] the archivelogs are restored to default archive destination. Prior to the restore you must ensure that there is space available for the restore of these archivelogs
[2] the archivelogs are restored from backup if they are not already on disk
[3] once recovery is completed RMAN will automatically remove them from disk
[4] time taken to recover this datafile
[5] date and time when recover completed


Alert.log

Only RMAN restore operations are written to the alert.log. User-managed restore sessions will not appear in the alert.log as they are performed outside of Oracle.


Tue Dec 27 14:05:14 EST 2011
Full restore complete of datafile 5 /opt/app/oracle/oradata/ORA102/example01.dbf.  Elapsed time: 0:00:07
  checkpoint is 2989194
  last deallocation scn is 399417



All recovery sessions, whether user-managed or RMAN will also be written to the alert.log. This is an example of a RMAN recovery session:

Tue Dec 27 14:05:55 EST 2011
alter database recover datafile list clear
Tue Dec 27 14:05:55 EST 2011Completed: alter database recover datafile list clear
Tue Dec 27 14:05:55 EST 2011
alter database recover if needed
 datafile 5
Media Recovery Start
 parallel recovery started with 2 processesORA-279 signalled during: alter database recover if needed
 datafile 5
...
Tue Dec 27 14:05:56 EST 2011
The input backup piece /opt/app/oracle/fra/ORA102/backupset/2011_12_27/o1_mf_annnn_TAG20111227T135926_7hlf4jrk_.bkp is in compressed format.Tue Dec 27 14:07:23 EST 2011 [1]
Archivelog restore complete. Elapsed time: 0:00:01 [2]
Archivelog restore complete. Elapsed time: 0:00:01
...
Tue Dec 27 14:07:31 EST 2011
alter database recover logfile '/opt/app/oracle/fra/ORA102/archivelog/2011_12_27/o1_mf_1_77_7hlfmdc7_.arc'
...
Tue Dec 27 14:07:31 EST 2011
Media Recovery Log /opt/app/oracle/fra/ORA102/archivelog/2011_12_27/o1_mf_1_77_7hlfmdc7_.arc
ORA-279 signalled during: alter database recover logfile '/opt/app/oracle/fra/ORA102/archivelog/2011_12_27 /o1_mf_1_77_7hlfmdc7_.arc' [3]
...
Tue Dec 27 14:07:31 EST 2011
Media Recovery Log /opt/app/oracle/fra/ORA102/archivelog/2011_12_27/o1_mf_1_87_7hlfmk96_.arc
Tue Dec 27 14:07:31 EST 2011
Recovery of Online Redo Log: Thread 1 Group 1 Seq 88 Reading mem 0  Mem# 0: /opt/app/oracle/oradata/ORA102/redo01.log [4]
Tue Dec 27 14:07:31 EST 2011
Recovery of Online Redo Log: Thread 1 Group 2 Seq 89 Reading mem 0
  Mem# 0: /opt/app/oracle/oradata/ORA102/redo02.log
Tue Dec 27 14:07:31 EST 2011
Media Recovery Complete (ORA102)[5]

[1] this was in fact a compressed backuppiece. This information is only displayed in the alert.log rather than the RMAN restore log
[2] the time taken to restore the archivelog
[3] ORA-279 is informational - confirming the archivelog required for recovery
[4] for complete recovery, Oracle will also need to apply the redo from the online logs
[5] end of recovery



User Managed Recovery Log

This is an example of a user-managed recovery. We are performing the recovery via SQL*Plus:



$ sqlplus

SQL*Plus: Release 10.2.0.5.0 - Production on Wed Dec 28 09:59:41 2011

Copyright (c) 1982, 2010, Oracle.  All Rights Reserved.

Enter user-name: / as sysdba

Connected to:
Oracle Database 10g Enterprise Edition Release 10.2.0.5.0 - 64bit Production
With the Partitioning, Data Mining and Real Application Testing options

SQL> alter database datafile 5 offline; [1]

Database altered.

SQL> recover datafile 5; [2]
ORA-00279: change 2989857 generated at 12/27/2011 12:50:00 needed for thread 1
ORA-00289: suggestion :
/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_79_7hnmjl2y_.arc
ORA-00280: change 2989857 for thread 1 is in sequence #79 [3]


Specify log: {<RET>=suggested | filename | AUTO | CANCEL}

ORA-00279: change 2989860 generated at 12/27/2011 12:50:01 needed for thread 1
ORA-00289: suggestion :
/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_80_7hnmjmbc_.arc
ORA-00280: change 2989860 for thread 1 is in sequence #80
ORA-00278: log file
'/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_79_7hnmjl2y_.arc' no
longer needed for this recovery [4]


Specify log: {<RET>=suggested | filename | AUTO | CANCEL}
auto [5]
ORA-00279: change 2989874 generated at 12/27/2011 12:50:39 needed for thread 1
ORA-00289: suggestion :
/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_81_7hnmjmkf_.arc
ORA-00280: change 2989874 for thread 1 is in sequence #81
ORA-00278: log file
'/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_80_7hnmjmbc_.arc' no
longer needed for this recovery

...
ORA-00279: change 2991001 generated at 12/27/2011 12:58:00 needed for thread 1
ORA-00289: suggestion :
/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_87_7hnmjoc7_.arc
ORA-00280: change 2991001 for thread 1 is in sequence #87
ORA-00278: log file
'/opt/app/oracle/fra/ORA102/archivelog/2011_12_28/o1_mf_1_86_7hnmjkm0_.arc' no
longer needed for this recovery


Log applied.
Media recovery complete. [6]
SQL>

SQL> alter database datafile 5 online; [7]

Database altered.

[1] take the datafile offline in preparation for a restore from your user-managed backup

[2] once the datafile is restored from your user-managed backup, recover it

[3] the first archivelog required to recover this file

[4] we have pressed the ENTER key, hence asking Oracle to apply the log requested

[5] if there are a lot of archivelogs to apply and they are all in the archive directory, use the AUTO option for Oracle to apply the rest of the required archivelogs. Otherwise you will need to manually specify each archivelog requested, or press ENTER when prompted for each archivelog

[6] recovery is now complete

[7] place the datafile online, thus making it available for use again



Operating System Utilities

The file being restored should be increasing in size until its actual size. The timestamp should also be changing as the file is being updated by Oracle. Use an operating system utility such as "ls -lt" to see this information.


$ ls -ltr <full path and file name being restored>

eg:
$ ls -ltr /database/db251/asbs/BLOB_DOC_IMAGES_B12.dbf


You can also use operating system utilities such as vmstat, sar and iostat to monitor resource utilization. Is the hardware working to its full capacity? Where is the bottleneck? Are there other I/O intensive operations happening on the host? Install Oracle's OSWatcher utility for more information if required.

Note: In Unix, if the job is started interactively from a terminal without NOHUP or CRON, ensure that the Unix job itself is still running. Sometimes your terminal may terminate due to idle timeouts and leave the job in an orphaned status. If the restore operation is going to take a long time you should consider using NOHUP or CRON to avoid this problem.


If the file is being restored to ASM you should also be able to monitor its file size in ASM. Utilities such as "ls -l" will work also work in ASM.

Please also be aware of this issue:
    Note 882555.1 - RMAN is Not Restoring OMF Datafiles in Their Original Location



Data Dictionary Views

All RMAN operations will have a corresponding database session. Therefore you can query the data dictionary to check its progress.

1) RMAN sessions

SQL> -- RMAN sessions

set linesize 100 trimspool on

COLUMN sid      FORMAT 9999
COLUMN serial#  ALIAS SER# FORMAT 99999
COLUMN spid  FORMAT 9999
COLUMN username FORMAT a10
COLUMN status   FORMAT a2
COLUMN program  FORMAT a32
COLUMN logon_time form a15
COLUMN module form a30
COLUMN action form a35
COLUMN process form a14

SELECT
       s.sid ,
       s.serial# "ser#",
       s.username,
       to_char(s.logon_time,'DD-MM-RR hh24:mi') logon_time,
       s.osuser,
       s.process,
       p.spid ,
       s.machine,
       substr(s.status,1,1) status,
       s.program
FROM v$session s, v$process p
WHERE upper(s.program) like '%RMAN%'
AND   s.paddr = p.addr (+)
ORDER by s.logon_time, s.sid
/


2) The percentage of work completed


Run the following queries at least 3 times, at 5 minute intervals to see progress/change.

SQL>set echo on feedback on
    column path format a50
    set header off
    select
           sl.sofar, sl.totalwork,
           round(sl.sofar/sl.totalwork*100,2) "% Complete"
    from   v$session_longops sl, v$session s, v$process p
    where  p.addr = s.paddr
    and    sl.sid=s.sid
    and    sl.serial#=s.serial#
    and    opname LIKE 'RMAN%'
    and    opname NOT LIKE '%aggregate%'
    and    totalwork != 0
    and    sofar <> totalwork;


3) Session Waits

Are there any sessions in wait and what is it waiting for?


Run the following queries at least 3 times, at 5 minute intervals to see progress/change.


set linesize 200 trimspool on

col event form a25
col p1text form a15
col p1 form 999999
col p2text form a15
col p2 form 999999
col p3text form a10
col p3 form 9999
col waited form 9999
col waiting form 9999

select sid, event, p1text, p1, p2text, p2, p3text, p3,
wait_time waited, seconds_in_wait waiting
from gv$session_wait
where event not like 'SQL*Net%'
and event not like '%timer%'
and event not like 'rdbms%'
and event not like 'pipe%'
and event not like 'DIAG%'
and event not like 'Streams AQ%'
and event not like 'VKTM%'
and state = 'WAITING'
order by seconds_in_wait
/



4) Recovery Progress


What is the recovery progress?  V$RECOVERY_PROGRESS is only populated when RECOVERY is in progress. A restore operation will not populate this view. So if you think a recovery process is slow - is it really at the recovery phase, or still restoring from RMAN backuppieces?

This is an example of a recovery progress:



22:27:38 SQL> select START_TIME,TYPE,ITEM,UNITS,SOFAR,TOTAL from v$recovery_progress;

START_TIME                  TYPE            ITEM                             UNITS                         SOFAR      TOTAL
--------------------------- --------------- -------------------------------- ------------------------ ---------- ----------
12-nov-14 16:08:10          Media Recovery  Average Apply Rate               KB/sec                        29713          0
12-nov-14 16:08:10          Media Recovery  Redo Applied                     Megabytes                    660747          0
12-nov-14 16:08:10          Media Recovery  Last Applied Redo                SCN+Time                          0          0
12-nov-14 11:28:16          Media Recovery  Checkpoint Time per Log          Seconds                           6          6
12-nov-14 11:28:16          Media Recovery  Standby Apply Lag                Seconds                           0          0







Redo recovery rate is determined by a number of factors:

1) PARALLEL_EXECUTION_MESSAGE_SIZE

The default value for this parameter may not be big enough, thus consider increasing to its maximum operating system dependent value:

SQL> show parameter PARALLEL_EXECUTION_MESSAGE_SIZE
SQL> alter system set parallel_execution_message_size=65535 scope=spfile;
This parameter change needs the database to be restarted/remounted.



2) native I/O rates at the hardware level - consult with your system administrator/hardware vendor for this

3) recovery parallelism

This is operating system dependent - Oracle will startup the required number of parallel processes to perform this task.

If you feel the need to manually specify this, then the commands are:

SQL> RECOVER datafile x,y,z parallel (degree 32);

OR

SQL> recover parallel 32;



4) if after tuning the above, redo apply rate is still not acceptable then you can temporarily set db_block_checking to false to try and boost recovery performance.


Media Management Logs


If restoring from tape, confirm that it is indeed restoring from tape rather than waiting on the media manager to service the request. Is the tape busy or idle? Ask your media management support team to confirm the rate at which data is being read from tape.