Dataguard replication Query

Hello there,
If I am using external Netapps storage and snapmirror to replicate data to my storage array at the DR site do i still need to use dataguard for log shipping?
I am a little confused where both fit. Are they basically both replication technologies, dataguard being at the database level, and netapps using storage level replication
Advice would be much appreciated
Cheers
C

Thanks Enrique and I guess that would be easier too... but...
I have spoken twice with them and no luck. They say that would be risky to DMZ, by opening direct communication between those servers (A and C) and bla bla... you know how huge those guys can be, so I think I'm not insisting them anymore... :)
Anyway, thanks for the response!
Charles.

Similar Messages

Architecture for APEX using iAS 10.1.2 and 10g with Dataguard replication

We are currently using APEX in a stand-alone architecture using iAS 10.1.2 on with 10g as the DB(all on the same server). Our APEX apps are fast becoming a significant part of our production environment and I would like to set up a redundant failover site using Dataguard replication. I am very familiar with the DB failover, but I am struggling with how I can set up iAS on the standby server.
Oracle guidelines for iAS failover using Dataguard (this is not AS Dataguard) specify that the primary and standby server local hostname must be identical for the failover architecture to work unless you use a virtual host name. Due to restrictions in our environment I am not able to set the lcoal hostname the same on the primary and standby servers. As an alternative I installed iAS(10.1.4) using a virtual hostname on my primary server, but it appears that the servername that shows up in all the installation references after the installation is still the actual server name, not the virtual name.
I am looking for guidance from someone that has successfully set up a iAS failover architecture using dataguard on servers that do not have the same local hostname.
thanks

Hi,
I have the same problem. Do you find something about putting jsf in OAS 10.1.2 ?
Regards

DataGuard Replication through a third server

Hey Oracle guys!
Hope you can help me with this or give some ideas...
I want to setup dataguard replication between "A" and "C" server. But "A" doesn't reach "C" server directly, instead, the connections to "C" must be done through a "B" server, ok? ... (A->B->C)... and also viceversa ("C" server cannot communicate to "A" directly, it needs "B")... I have spoken to the network admins and they said: "that's the current architecture and can't be changed by now"...
At the beginning I was having some ideas, about installing Oracle on "B" server and trying to "cheat" Oracle by starting a Listener pointing to "C" server (on host section, in the listener.ora) and configure the primary's server standby-destination to this "B" fake listener... but that obviously is not going to work...
Do you have some ideas guys? or do you think that's possible?
Thanks and cheers!
Charles

Thanks Enrique and I guess that would be easier too... but...
I have spoken twice with them and no luck. They say that would be risky to DMZ, by opening direct communication between those servers (A and C) and bla bla... you know how huge those guys can be, so I think I'm not insisting them anymore... :)
Anyway, thanks for the response!
Charles.

ACS server replication Query

Hi All ,
I have two ACS server primary & secondary server . New secondary server to be deployed into network . My primary ACS server has got 1000 AAA clients configured with 15000 user id configured in multiple group profile . My question over here is when i do database replication between primary and secondary ,whether entire databse will be replicated from my primary server to secondary server like all AAA clients and end user , group profile , interface configuation etc , else it will replication has got restriction for database .
Totally : AAA clients & User ID will be on one database backup or it will reside on differnt location
kindly clarify me over here ,Thank you .

Hi,
The entire Database will get over written in case of database restore.
You use ACS Database Replication to copy various components of the ACS internal database to other ACSs. This method can help you plan a failover AAA architecture, and reduce the complexity of your configuration and maintenance tasks.
The components that can be replicated are:
User and group database
Group database only
Network Configuration Device tables
Distribution table
Interface configuration
Interface security settings
Password validation settings
EAP-FAST master keys and policies
Network Access Profiles
Logging Configuration (Enable/Disable Settings)
The following link will give you details of the database replication.
http://www.cisco.com/en/US/docs/net_mgmt/cisco_secure_access_control_server_for_windows/4.2/user/guide/SCAdv.html#wp756304
Hope this helps.
Regards,
Anisha
P.S.: Please mark this thread as resolved if you feel your query is resolved. do rate helpful posts.

Dataguard Replication Encryption

Hello all,
We plan to use Dataguard on Oracle 9.2.0.8 version, running in an AIX 5.2 platform. We want the traffic between primary and secondary database to be encrypted, but Oracle Advance Security Option is too expensive for us.
Is any other alternative method be used if we just want the encryption?
a. Is the VPN IPSEC tunnel setup between primary and secondary server (VPN IPSEC tunnel setup between 2 AIX) a workable solution?
b. Is SSH tunnel another workable solution?
Thanks

I think you need to address this with Unix/AIX forums becuase it now outside of Oracle domain. You are nolonger talking about Dataguard and ASO.

Golden Gate replication query

Hi,
Is it possible to transform one row from source database into multiple rows in target database using Golden Gate?
Thanks,
Deepak

Another way is using SQLEXEC:
Suppose we have:
at source db -
create table test.wide (id number primary key, COL1 VARCHAR2(25), COL2 VARCHAR2(25), COL3 VARCHAR2(25));at target db -
create table test.thin (id number primary key, col_name VARCHAR2(10), col_value VARCHAR2(25));then in replicat params -
--Replicat group --
REPLICAT REP2
--source and target definitions
--ASSUMETARGETDEFS
SOURCEDEFS /u01/app/oracle/product/11.1.1.1.2ogg411g/dirdef/sourcedef
--target database login --
USERID ogg, PASSWORD ogg
--file for dicarded transaction --
DISCARDFILE /u01/app/oracle/product/11.1.1.1.2ogg411g/discard/rep1_discard.txt, APPEND, MEGABYTES 10
--ddl support
DDL
--Specify table mapping ---
MAP test.t1, TARGET test.t11g, COLMAP (USEDEFAULTS, VCH=@STRNUM(NM));
MAP test.CITIES, TARGET test.CITIES;
MAP test.COUNTRIES, TARGET test.COUNTRIES;
MAP test.LOBEXP, TARGET test.LOBEXP;
MAP test.wide, TARGET test.thin, COLMAP (USEDEFAULTS, COL_NAME="COL1", COL_VALUE=COL1), &
SQLEXEC (ID COL2PROC, QUERY "insert into test.thin values (:id_params+1,'COL2',:value_param)", PARAMS (id_params = id, value_param = COL2)), &
SQLEXEC (ID COL3PROC, QUERY "insert into test.thin values (:id_params+2,'COL3',:value_param)", PARAMS (id_params = id, value_param = COL3)), &
SQLEXEC (ID COL_DELE, ON DELETE, QUERY "delete from test.thin where id in (:id_params+1,:id_params+2)", PARAMS (id_params = id)), &
SQLEXEC (ID COL2UPDA, ON UPDATE, QUERY "update test.thin set col_value=:value_params where 1=:is_updated and id=:id_params+1", PARAMS (is_updated = @IF(@COLTEST(COL2,MISSING,
INVALID),0,1), value_params = COL2, id_params = id)), &
SQLEXEC (ID COL3UPDA, ON UPDATE, QUERY "update test.thin set col_value=:value_params where 1=:is_updated and id=:id_params+2", PARAMS (is_updated = @IF(@COLTEST(COL3,MISSING,
INVALID),0,1), value_params = COL3, id_params = id));
--MAP test.account, TARGET test.thin, COLMAP(USEDEFAULTS, COL_NAME='COL1', COL_VALUE=COL1);
DDLERROR 24344 DISCARD;An extractor params would be -
--extract group--
EXTRACT ext1
--connection to database--
USERID ogg, PASSWORD xxx
EXTTRAIL /u01/app/oracle/product/11.1.1.12ogg/dirdat/ss
SEQUENCE test.*
--DDL support
DDL INCLUDE MAPPED OBJNAME test.*
--DML
TABLE test.*;Works fine - just tested.
Edited by: Artem Khisamiev on 15.10.2012 4:09
Add DELETE and UPDATE DML improvements to replicat params.

Dataguard replication

hello oracle experts
I would like to share with you a brave scenario which it would be perfect if I could implement it
SCENARIO
I have a 9.2.0.7 production database which is in SUN solaris with VERITAS cluster not ORACLE RAC.
i would like to replicate - create a physical standby databse in an 11gR2 environment with linux RED HAT platform and ORACLE ASM CLUStER
is that POSSIBLE?
Can I use also Golden Gate for this or with Data Guard I am fine
Thanks in advance
Edited by: airmax012 on 22 Σεπ 2011 2:48 πμ

airmax012 wrote:
I forgot to tell that the new physical standby database I would like to be in 11gR2 version and not 9.2.0.7Standby database must have same version.So you can't configure standby where your primary is on version 9.2.0.7 and secondary database on 11gR2.
Hope this help you
--neeraj

9.2.0.6 DataGuard - how to "break" replication, make standby "standalone"

Hi All,
I have an existing 9.2.0.6.0 DataGuard environment, where currently I just have one primary and one physical standby database. I am planning to add in another physical standby database, but then at some point, I want to "break" the dataguard replication to this one, and basically take it out of the DataGuard environment, and have it run as a standalone database. I am not talking about putting it into "open read-only" mode, I want to completely remove it from the DataGuard replication, and have it up and running as a standalone db. Is that even possible? I don't see anything mentioned about this in the official DataGuard documentation.
If anyone has any thoughts on the best way to do this, I'd really appreciate it!!
Thanks,
Brad

Actually, both... here is my potential plan:
1. Put a copy of the production database onto the new hardware, and instantiate it into the DataGuard configuration as a new physical standby database.
2. "Break" the dataguard replication, so that the new database on the new hardware becomes a standalone database, separate from DataGuard. At this point, we can test out some of the application against this database. The testing could last up to a few weeks.
3. Once we are happy with the testing, then we need to re-create this database as a backup from production, and get it back into the DataGuard replication as a standby. When it is time to cutover the production system, then we do a DataGuard Switchover, so that the database on the new hardware becomes the new primary database, replicating out.
Does that sound feasible? I think I can take care of all of those steps, the only one that I don't really know how to do is step 2, where we "break" the dataguard replication, and bring up the database on the new hardware as a fully stand-alone db.
Thanks,
Brad

Dataguard Configuration under Juniper Firewall

Dear Friends ,
We are using Oracle database 11g with Oracle Ent Linux 5.8. We are using active dataguard in this system . Recently we are deployed firewall both on PRIMARY and STANDBY end . we are using juniper firewall device . After deployment of Juniper Firewall on STANDBY end , we observe that Dataguard replication is not performed , i mean log of the PRIMARY server does not transmitted to the STANDBY end .
Does is there any recommendation using firewall device with DATAGUARD ?
How can I resolve this Dataguard replication problem under the Firewall ?

The firewall simply placed on top of STANDBY database . nothing is configured in firewall . I mean all ports are open and no rules are applied yet .
Before firewall , redo transmitted successfully but after placing the firewall , it is not working .
It means you agree the firewall is blocking the transmission and it also means that ports are blocked.
On top of what have been said -
Check the connectivity to Standby from Primary as SYSDBA ?
Check the Primary database & Standby database alert log ?
Pradeep

Listener not working in one of the servers of the Standby site

Hello,
The system is a RAC with two servers working as Primary site, and two servers working as Standby site.
Both servers of the Primary site are sending redo logs to one of the servers of the Standby site. The other server of the Standby site is not receiving any redo log.
I have checked the listener.ora and the tnsnames.ora files of the server that is not receiving and it is well configured comparing it with the server that is receiving the redo logs.
If I check the status of the listener: lsnrctl status listener_<node name>
Both of the Standby servers look fine.
But if I connect directly to the database...
In the 1st server of the Standby, the IP appears:
SQL> show parameter listener
NAME TYPE VALUE
local_listener string (ADDRESS = (PROTOCOL = TCP)(HOST = xxx.xxx.xxx.xxx)(PORT = xxxx))
remote_listener string LISTENERS_DB
Nevertheless, in the second server of the Standby site, the one that is not receiving the redo logs, the IP does not appear:
SQL> show parameter listener
NAME TYPE VALUE
local_listener string (ADDRESS = (PROTOCOL = TCP)(HOST = )(PORT = xxxx))
remote_listener string LISTENERS_DB
I have found the following to reload the listener:
Oracle# lsnrctl reload listener_`hostname`
Oracle# lsnrctl reload listener_standby
Has anyone tried this? Did it work? It can be run without affecting the sytem, Dataguard, replication..., right?
If you have another idea to fix this, let me know please.
Thank you ;)
Edited by: user11281526 on 29-oct-2009 4:35
Edited by: user11281526 on 30-oct-2009 7:36
Edited by: user11281526 on 31-oct-2009 4:11

If I see the status of the listener_standby in the second server of the Standby site, the one that it is not receiving redo logs, it looks like it is fine. Why?
#lsnrctl status listener_standby
LSNRCTL for Solaris: Version 10.1.0.4.0 - Production on dd-mm-yyyy hh:mm:ss
Copyright (c) 1991, 2004, Oracle. All rights reserved.
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=IP_second_server_standby)(PORT=xxxx)))
STATUS of the LISTENER
Alias listener_standby
Version TNSLSNR for Solaris: Version 10.1.0.4.0 - Production
Start Date dd-mm-yyyy hh:mm:ss
Uptime xx days yy hr. zz min.
Trace Level off
Security ON: Local OS Authentication
SNMP OFF
Listener Parameter File /opt/oracle/product/10.1.0/db_1/network/admin/listener.ora
Listener Log File /opt/oracle/product/10.1.0/db_1/network/log/listener_standby.log
Listening Endpoints Summary...
(DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=IP_second_server_standby)(PORT=xxxx)))
Services Summary...
Service "DB1" has 1 instance(s).
Instance "DB1", status UNKNOWN, has 1 handler(s) for this service...
The command completed successfully
Edited by: user11281526 on 29-oct-2009 6:41
Edited by: user11281526 on 29-oct-2009 6:42
Edited by: user11281526 on 29-oct-2009 6:43

Compare tables, find rows that are "different"

A common problem in replication/query-extract-load scenarios is comparing two sets of data, and finding the rows that have changed or are different. A clean solution to this is less obvious than it sounds, because "different" can involve nulls (love them!/hate them!).
Consider two structurally identical tables with all data linked by a common PK:
Snapshot A Snapshot B
PK First Last State PK First Last State
1 Bob Smith AZ 1 Bob Smith AZ
2 Louise Jones FL 2 Louise Brown FL
3 Joe Jones NULL 3 Joe Jones CA
4 Joe NULL GA 4 Joe Celko NULL
5 Phill NULL AL 5 Phill NULL AL
If we are interested in tracking or processing any changes in our data, we would agree that only Rows 1 and 5 are "the same" in both sets A & B.
Because of null logic, this sql only gets me row 2 (besides being syntactically miserable with a wide table) :
select A.PK from A inner join B on A.PK = B.PK
where (A.First <> B.First) OR (A.Last <> B.Last) OR (A.State <> B.State);
So how can I select rows 2, 3, and 4? (Extra credit for for a query I can type in less than two minutes with a pair of tables that have 42 columns)
Thanks much,
Steve Pence
DBA
Wycliffe Bible Translators
Orlando
[email protected]

Sorry, my post got mangled by the white space deleterer...
A common problem in replication/query-extract-load scenarios is comparing two sets of data, and finding the rows that have changed or are different. A clean solution to this is less obvious than it sounds, because "different" can involve nulls (love them!/hate them!).
Consider two structurally identical tables with all data linked by a common PK:
Snapshot A
PK First Last State
1 Bob Smith AZ
2 Louise Jones FL
3 Joe Jones NULL
4 Joe NULL GA
5 Phill NULL AL
Snapshot B
PK First Last State
1 Bob Smith AZ
2 Louise Brown FL
3 Joe Jones CA
4 Joe Celko NULL
5 Phill NULL AL
If we are interested in tracking or processing any changes in our data, we would agree that only Rows 1 and 5 are "the same" in both sets A & B.
Because of null logic, this sql only gets me row 2 (besides being syntactically miserable with a wide table) :
select A.PK from A inner join B on A.PK = B.PK
where (A.First <> B.First) OR (A.Last <> B.Last) OR (A.State <> B.State);
So how can I select rows 2, 3, and 4? (Extra credit for a query I can type in less than two minutes with a pair of tables that have 42 columns)
Thanks much,
Steve Pence
DBA
Wycliffe Bible Translators
Orlando
[email protected]

Cisco Unity Connection cluster issue

Hello,
I have a Cisco Untiy Connection cluster of two servers, one publisher and one subscriber. They are running software: 7.1.3ES43.33034-43. Yesterday at about 8:43 I lost connection to the subscriber server. I could still ping it but was unable to get to it via GUI nor SSH. I spoke to TAC yesterday and they told me to simply reboot the server. I was a little concerned about this because I don't want to cause some kind of split-brain effect. I work in a hospital and it is imperative that the voicemail system stay up. My only other option would be to do this at 4:00 am on a Sunday. Any suggestions?
Thanks

Hi David and all,
we have a unity connection 9.x pair. after an IP address change on the subscriber we are getting the error message:
‘Communication is not functioning correctly between the servers in the Cisco Unity Connection cluster’
Both servers respectively show its own status as a publisher but lost communication with the other server (if we check cluster management from either server).
show tech network hosts – displays correct info.
show perf query class "Number of Replicates Created and State of Replication"
==>query class :
- Perf class (Number of Replicates Created and State of Replication) has instances and values:
ReplicateCount -> Number of Replicates Created = 603
ReplicateCount -> Replicate_State = 2
On both servers the replication is showing as 2 – so its good.
Various reboots have been done, services are up and working and there are no firewalls between server’s .
Any advice appreciated.
TIA…Jeff

Issue with User web page into Call Manager

Hello
I have an issue where all my users can log into their User page on UCM but when they make a change to speed dial or anything else, the change does not reflect on the phone in UCM. Doing a reset from their web page doesn't reset the phone. This is UCM 7.1.5 and the standard ccm end user group is enable for all end users.
thanks

Hi Bill -
Have you verified your User Management Roles and effective permissions? Go to User Management - Roles - select Standard CCMUSER Administration and for the permissions you want, ensure both Read and Update are selected.
Role Information
Application
Cisco Call Manager End User
Name
Description
Resource Access Information
">Resource Description Privilege
read                                               update
CCMUser: Access List
read                       update
CCMUser: Device
read                       update
CCMUser: Directory
read                       update
CCMUser: Fast Dials
read                       update
CCMUser: IP Phone Services
read                       update
CCMUser: Line Settings
read                       update
CCMUser: Personal Address Book
read                       update
CCMUser: Plugins
read                       update
CCMUser: RemoteDestination
read                       update
CCMUser: Service URL
read                       update
CCMUser: Speed Dial User
read                       update
CCMUser: User Settings
read                       update
You can also check an individual's effective permissions by using these procedures:
Viewing a User's Roles, User Groups, and Permissions
This section describes how to view the roles, user groups, and permissions that are assigned to a user that belongs to a specified user group. Use the next procedure to view the roles, user groups, and permissions that are assigned to a user in a user group.
Note: You can also view user roles by using User Management > Application User (for application users) or User Management >End User (for end users) to view a particular user and then display the user roles.
Choose User Management > User Group.
The Find and List User Groups window displays.
Find the user group that has the users for which you want to display assigned roles.
Click the name of the user group for which you want to view the roles that are assigned to the users.
The User Group Configuration window displays for the user group that you chose. The Users in Group pane shows the users that belong to the user group.
For a particular user, click the i icon in the Permission column for the user.
The User Privilege window displays. For the user that you chose, this information displays:
User groups to which the user belongs
Roles that are assigned to the user
Resources to which the user has access. For each resource, this information displays:
Application
Resource
Permission (read and/or update)
Now if both items above look OK, you might check your DB replication status. I assume you have a Publisher and one or more Subscribers? User phones registered to Subscriber? You can check replication several ways:
CLI
RTMT
Unified Reporting on CUCM administrator web page - select "Unified CM Database Replication Debug" report. This is the easiest.
The desired Replication State is 2.
Here is some further information:
Check the DB replication status on all the Cisco Unified Communications Manager nodes in the cluster to ensure that all servers are replicating database changes successfully. You can check by using either RTMT or a CLI command.
• To check by using RTMT, access the Database Summary and inspect the replication status.
• To check by using the CLI, enter the command that is shown in the following example: admin: show perf query class "Number of Replicates Created and State of Replication"
==>query class :
- Perf class (Number of Replicates Created and State of Replication) has instances and values:
ReplicateCount -> Number of Replicates Created = 344 ReplicateCount -> Replicate_State = 2
Be aware that the Replicate_State object shows a value of 2 in this case. The following list shows the possible values for Replicate_State:
0—Replication Not Started. Either no subscribers exist, or the Database Layer Monitor service is not running and has not been running since the subscriber was installed.
1—Replicates have been created, but their count is incorrect.
2—Replication is good.
3—Replication is bad in the cluster.
4—Replication setup did not succeed.
I'm thinking it's more your Roles/permissions and not replication, but I included just in case. Hope this helps!
Ginger

Are these independent or go together

I am the senior dba on our project and currently working on implementing high availability practices/procedures for our database. Currently we don't use any add on products and basically we run 9-5 and do backups at night. We are interested in using some of the technologies such as DATAGUARD, REPLICATION, ORACLE STREAMS, FLASHBACK TECHNOLOGY. We also want to create a better disaster recover plan.
From a high level, the question I have is what technologies go-together and what are not related.
Example, If I use DataGuard, Will I need to also use replication? Or does Dataguard handle the same kind of things. What/how do I get flashback technology?
Can anyone recommend the best approach for high availability databases? 24x7
I'm really just looking for high level answers and I'll follow up with some research.
Thanks for your time,
Mike O.

If you are going to use 10g RAC, then you would only need Dataguard in a recovery situation (e.g.: you entire building burned down and the 2+ RAC nodes were both in the building along with your backup tapes, etc)....
If you only want a single node and want to have an offsite server that is ready to go if your building was destroyed (for whatever reason), then dataguard is what you want. If you want to be able to failover and load balance, then RAC is what you want, but RAC (and someone may disagree) can't be run in two different datacenters (unless they were VERY close together and connected by dart fiber or something.
I've not implemented dataguard (but have implemented standby server (which is what dataguard is now)), but be aware, not everything is shipped to the dataguard server (e.g.: if a user updates, inserts, etc, no logging, you just lost that data, but perhaps dataguard has fixed this issue? I know standby server did not).
-Greg

Backup of copy while recovery of copy running

Have a question on an RMAN error we are getting and wondering if anyone else has done anything like this and seen this error.
We have had an SR open since 8/30 with no resolution yet though the SR is a standard severity as this problem is not severe and we can work around it. Our opinion is that what we are trying to do should work without error.
Environment Details. I do not suspect this matters as I have reproduced this on Solaris and linux as well as DB versions 11.2.0.3, 10.2.0.4 and 10.2.0.5
Error:
     channel ORA_DISK_1: finished piece 1 at 25-SEP-12
     piece handle=/oracle/RMAN_FULL_SEG2/FULL_SEG2_16_1_794915068 tag=SCOTTSFULL comment=NONE
     channel ORA_DISK_1: backup set complete, elapsed time: 00:01:42
     channel ORA_DISK_1: starting incremental level 0 datafile backup set
     RMAN-00571: ===========================================================
     RMAN-00569: =============== ERROR MESSAGE STACK FOLLOWS ===============
     RMAN-00571: ===========================================================
     RMAN-03009: failure of backup command on ORA_DISK_1 channel at 09/25/2012 09:46:10
     ORA-19588: datafile copy RECID 2 STAMP 794914328 is no longer validHope this makes sense. It's a bit complex to explain.
Steps to Reproduce some of these can be skipped I was just documenting
everything I did for the SR to reproduce the issue.
*1) Create simple database via DBCA*
Create directories to hold the copy and the full backup.
I used the following directories.
          /oracle/oradata for the database
          /oracle/RMAN for the copy
          /oracle/RMAN_FULL_SID for the backup
          /oracle/IRMAN for incremental that is later applied to DB
*2) Create a copy of the database into /oracle/RMAN directory*
     backup incremental level 0 as copy db_file_name_convert=('/oracle','/oracle/RMAN') reuse
database tag="SCOTTTEST"
     verify copy with 'list copy of database' if you want to
*3) Do backup just to establish a baseline for incremental. (this one not really relevant)*
     backup incremental level 0 tag "SCOTTSFULL" fORMAT
'/oracle/RMAN_FULL_SEG2/FULL_%d_%s_%p_%t' copy of database;
*4) Do incremental backup that will be applied to copy.*
     backup incremental level 1 FORMAT '/oracle/IRMAN/incremental_refresh_%U' for
recover of copy with tag "SCOTTTEST" database;
Here is where issue happens.
Start 2 sessions into rman at this point
one for number 5 and one for number 6.
Start number 5 running and wait for the channel ORA_DISK_1: starting piece message.. then run #6
*5) Start backup of that copy with filesperset 1 to force multiple backupsets.*
The filesperset clause is not set to 1 in our production databases but the backups
are much bigger there and obviously take longer. The use of filesperset is just
to reproduce the issue on a much smaller database.
This has to take longer than #6 and there has to be a channel finishing a set of files and
starting on a new file after #5 is complete.
     backup incremental level 0 filesperset 1 tag "SCOTTSFULL" fORMAT
'/oracle/RMAN_FULL_SEG2/FULL_%d_%s_%p_%t' copy of database;
*6) Start refresh of copy after starting number 5 above.*
     RECOVER COPY OF DATABASE WITH TAG "SCOTTTEST";
# 6 Will run quickly. When #5 finishes a file and moves to the next file it will error.
There are some work arounds that do work and avoid the error.
1) Set filesperset large enough to cover all the files in
the DB (in above example change filesperset to 15 no error will be reported)
2) allocate enough channels to ensure that you cover all the files in the database
3) Don't run the recover of the copy while a full backup is running.
4) Don't run the backup on the copy but backup the original DB.
The situation in our production DB is this.
We want to recover the copy every 3 hours keeping ourselves no longer than 3 hours out of date
with the main database.
every couple of days we do an incremental level 0 of the copy which on our bigger databases
can take more than 3 hours.
These level 0's fail if they run longer than 3 hours.
The answers on the SR so far have been to increase filesperset to a huge number and don't
run them at the same time. The reason they are giving us is that we are not getting a consistent backup
because the files are being recovered while the backup runs.
The problem I have with that reasoning is that if you choose options 1 and 2 above that do stop the error
the recovery is still running while the backup is and the files are still changing so in that case the
files are not consistent either but no error is thrown. Am I thinking about this completely wrong?
We feel that those are not good solutions with what we are trying to do and wonder if anyone
else has tried to do this type of thing with any success?
My suspicion on what is happening is that during the start of the level 0 backup RMAN grabs a list
of files with their id's to backup. The recovery actually marks files deleted and adds a new file with
a new id in the controlfile/catalog metadata (I forget the table at the moment) . Once RMAN finishes a
set of files/file it tries to move on to the rest and cannot find a file with the same id that is
not deleted anymore so throws the not valid error.

We want to recover the copy every 3 hours keeping ourselves no longer than 3 hours out of date
with the main database.Sometimes you need to redesign your strategy. If the above line is your target, you should consider other solutions such as Active/Standby Dataguard replication. Why go through backup and recovery on a remote site, when you can ship the archivelogs to remote destination and have the logs applied, thus keeping both sites up-to-date.
Following is a link for Active/Standby:
http://www.oracle.com/technetwork/database/features/availability/twp-dataguard-11gr2-1-131981.pdf

Dataguard replication Query

Similar Messages

Maybe you are looking for