Thanks to everyone for your help. I finally was able to complete the RMAN
disaster recovery! I am backing up to NFS-mounted disk, which is
subsequently copied to tape. Oracle 8.1.6 on Compaq Tru64. Here is what I
learned:
1. Even if you use the RMAN catalog to back up your database, you can
recover the database from the control file alone. But I wouldn't set the
parameter that clears RMAN data from the control file really low.
2. Just because you write RMAN backup files to an NFS-mounted drive doesn't
mean RMAN will be able to recover them. That was the bottom line on my
"stuck" symptoms. Eventually Oracle support "suggested" that I take NFS out
of the picture. I moved the files to local disk and recovery worked fine.
Since then my sys admin pushed harder with the system vendor, got further
suggestions for things like IP threads, and I am recovering successfully
from NFS.
3. You will need the archive log files from around the time of the backup.
Evidently when RMAN is doing its online backup, it is counting on using the
data in the redo logs that are being produced while the backup is in
progress to complete the recovery. If you are using RMAN to back up your
archive logs, this isn't an issue. In my case, it seemed to just be doubling
the disk space needed for storing archive logs online.
4. Thanks to everyone for the date format suggestions. The one that finally
worked was:
set until time "to_date('082620021229','mmddyyyyhh24mi')";
5. Even after applying the archive logs, RMAN was still not able to complete
the recovery. It was probably wanting the active redo logs from the time of
the backup. I exited RMAN and fired up svrmgrl and issued "alter database
open resetlogs" and I was home free. Recovery was complete. In a true
disaster recovery situation where you only have the backup tape, I'm not
going to quibble about the last transaction or so.
Again, thanks to everyone for their help, and I hope others can learn
something from my experiences.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-----Original Message-----
Sent: Friday, August 16, 2002 12:05 PM
To: '[EMAIL PROTECTED]'
Thanks Jim. This morning I did something close to what you describe. I just
did
run {
allocate channel d1 type disk;
restore database;
}
Same result. It recovers the same set of files and hangs. At least this
rules out a lot of things like the time format of time. Now I'm thinking
maybe I should kill this RMAN session and issue it again. Since it reports
successful recovery of these files, maybe it will go past them and recover
more of the files. Thanks to you and everyone for their great suggestions.
As long as I have ideas I can keep plugging away at this.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-----Original Message-----
Sent: Friday, August 16, 2002 9:39 AM
To: Multiple recipients of list ORACLE-L
Dennis,
Try this:
sql> startup mount;
sql> exit
rman target sys/password nocatalog
then,
run {
allocate channel d1 type disk;
restore database;
recover database until cancel;
alter database open resetlogs;
}
Is you're controlfile coming from the RMAN backup set ora are you
copying it from the production box?
I discovered the control file is backed-up during an RMAN Level 0 at
the beginning of the process before the datafiles, therefore the
backed-up control file doesn't know about the backup set being run ( at
least this was the case during a database clone attempt).
...JIM...
>>> [EMAIL PROTECTED] 8/15/02 6:29:16 PM >>>
Oops, I forgot to clarify that I have the production database in
archivelog
mode, but the recovery database not in archivelog mode.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-----Original Message-----
Sent: Thursday, August 15, 2002 5:08 PM
To: '[EMAIL PROTECTED]'
James
I think you may have put your finger on a possible misconception of
mine.
Here is my situation/understanding.
- On production,
- Archive logging.
- RMAN backup to disk without shutting the database down.
- Not using RMAN to backup the archive logs.
- Disaster recovery scenario.
- This is a "burn the server" scenario. Imagine the computer
room no
longer exists. All you have is the backup tape from the offsite
storage. No
stringent recovery timeframe. If you tell the managers that it will
take you
a week to recover the data, no big deal. If you tell them you cannot
recover
the data because you forgot to copy some critical file to tape, that is
a
big deal.
- I would assume that all of you that use RMAN have performed
such a
test.
- My concept was to perform the equivalent of a cold
backup/cold
recovery. Just recover using the RMAN backup set that was written to
disk
and subsequently written to tape.
- What RMAN commands should I use to perform this recovery? I
have
made assumptions, but they may not be correct.
- I assumed this would be an incomplete recovery since I can't
recover to the present time. So I inserted the SET UNTIL TIME command
because I think that is how you get RMAN to perform an incomplete
recovery.
I picked a time just after the end of the RMAN backup.
- Do I need any archive logs under this scenario? I can change
my
production procedure to use RMAN to backup the archive logs if that is
what
is required.
- Do I need the RECOVER DATABASE command?
Any ideas would be appreciated. Thanks for everyone's patience while I
flail
around with this.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-----Original Message-----
Sent: Thursday, August 15, 2002 4:08 PM
To: Multiple recipients of list ORACLE-L
Dennis,
This is just a wild guess and I'm probably wrong but I saw in you're
original post this DB was not in archivelog mode, try putting it in
archivelog mode and running the restore again???
...JIM...
>>> [EMAIL PROTECTED] 8/15/02 2:58:31 PM >>>
Okay, I implemented everyone's comments and re-executed the RMAN
recovery.
Here is what I did and the results.
1. Action: Removed "alter database open resetlogs" from the run
statement.
Result: No change.
2. Action: Added trace=1 to the allocate channel command.
Result: No trace file is produced in udump.
3. Action: Reviewed Note 145624.1
Result: Did not see the solution to my problem. Most of the
suggestions
seem appropriate to backup rather than recovery jobs.
4. Action: Added log and debug trace statements to rman invocation
line.
Result: Produced log and trace file. The trace file at the point of
the
recovery hang contains the following:
krmxrpc: xc=5372006336 kpurpc2 returned 0
krmxrpc: xc=5372006336 RPC #10 completed immediately
RMAN-08523: restoring datafile 00006 to /ora05/ams/data0501.dbf
krmxrpc: xc=5372006336 kpurpc2 returned 0
krmxrpc: xc=5372006336 RPC #11 completed immediately
RMAN-08523: restoring datafile 00017 to /ora05/ams/rbs01.dbf
krmxrpc: xc=5372006336 kpurpc2 returned 0
krmxrpc: xc=5372006336 RPC #12 completed immediately
RMAN-08523: restoring datafile 00025 to /ora05/ams/mls_data0401.dbf
krmxrpc: xc=5372006336 kpurpc2 returned 3123
krmxrpc: xc=5372006336 starting longrunning RPC #13 to target:
DBMS_BACKUP_RESTO
RE.RESTOREBACKUPPIECE
krmxr: xc=5372006336 started long running rpc
krmxpoq: xc=5372006336, action="0000013 STARTED", col_l=15, ind=0,
sid=13
krmxr: callback returned TRUE, skipping sleep
krmxpoq: xc=5372006336, action="0000013 STARTED", col_l=15, ind=0,
sid=13
krmxr: sleeping for 1 seconds
krmxpoq: xc=5372006336, action="0000013 STARTED", col_l=15, ind=0,
sid=13
krmxr: sleeping for 2 seconds
krmxpoq: xc=5372006336, action="0000013 STARTED", col_l=15, ind=0,
sid=13
krmxr: sleeping for 4 seconds
krmxpoq: xc=5372006336, action="0000013 STARTED", col_l=15, ind=0,
sid=13
krmxr: sleeping for 8 seconds
And the trace continues with this statement.
Any suggestions would be appreciated, as are the suggestions to this
point.
I'm about ready to think it is TAR time.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-----Original Message-----
Sent: Wednesday, August 14, 2002 2:23 PM
To: '[EMAIL PROTECTED]'
I am trying to perform an RMAN disaster recovery task. While I use an
RMAN
catalog to make backups, I am trying to recover using just the control
file
information.
Oracle 8.1.6, Compaq/HP Tru64
I start RMAN with
rman target sys/password nocatalog
then,
startup mount
run {
set until time "to_date('08/11/2002 01:00:00','MM/DD/YYYY
HH24:MI:SS')";
allocate channel d1 type disk;
restore database;
recover database;
alter database open resetlogs;
}
Everything appears normal for awhile. In the alert log RMAN tries to
find
each file, doesn't find them. Then it successfully recovers 5 data
files
(including system and rollback) and reports success in the alert log.
Then .
. . nothing for hours. RMAN doesn't return an error. The RMAN shadow
processes are still present but with no CPU consumption. Nothing is
written
to the alert log.
I check V$SESSION_WAIT, and the only entry for the RMAN shadow
processes is one is SQL*Net message to client with seconds_in_wait =
0,
state = waited unknown time.
In V$SYSTEM_EVENT, time_waited and average_wait are zero for all
events. The following events have values of total_waits that are
increasing:
Increase in total_waits in
10-minutes
rdbms ipc message 401
pmon timer 57
control file parallel write 56
SQL*Net message to client 24
SQL*Net message from client 24
virtual circuit status 5
dispatch timer 3
smon timer 1
Archiving is turned off.
I have attempted this recovery many times using different RMAN backup
sets,
but the system always hangs at this point.
Any ideas would be appreciated.
Dennis Williams
DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: DENNIS WILLIAMS
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: James Howerton
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: DENNIS WILLIAMS
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: James Howerton
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: DENNIS WILLIAMS
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
