Multimaster replication survival & overhead

Hi
We are planning a system which should have a very high survivability rate. To this aim we consider using 3 oracle servers (8.1.7), located in 3 different sites, replicated in an asynchronous multimaster configuration, with a short refresh time. The nature of the application is that many small update transactions are happening constantly.
Several questions:
1. We are trying to assess the required bandwidth of the site-to-site network connection we will need to support this configuration. What is the approx. network overhead of a small (~500 bytes) transaction?
2. Provided we come up with a smart enough conflict resolving mechanism, will this system survive? It is vital that committed transactions will never, ever, be lost.
3. Can the 3 db architecture can be left in an inconsistent state? For example, can it be that a committed transaction in database A, is propagated to database B, and then A fails before it can be propagated to database C? leaving inconsistencies in working databases B & C (atleast until A is restored).
4. How difficult would be to return such a system to normal state after one (or more) sites fails?
Thanks,
Idan

<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by Patricia McElroy ([email protected]):
This requires a bit more detail - what indicates that replication is hung... is there anything in any of the database logs that indicates a problem. You indicate that even changes to the local database can not be made - is this true also for non-replicated tables? Also, Run "Select * from dba_repprop;" and confirm that the method for propagation is indeed ASYNCHRONOUS. <HR></BLOCKQUOTE>
thanks patricia
here is the details:
1. i am having 2 DB servers at remote locations running oracle 8.1.6 on solaris 2.7
2. When thre is a link fails between these two server, i am not able to put the job in a deffered que transactions though i have scheduled job and link as well.
3. it's not allowing me to commit insert statement on local server,when the link is down.
null

Similar Messages

Comparision between Multimaster replication and data guard

Hi,
I have some questions regarding Multimaster Replication and Data Guard. Like
*1.)* I have a web site having database in oracle 10.2.0 and multimaster replication configured on it. But whenever i need to change the structure of the tables i
need to stop replication. So my web site is unavailable but my first priority is availability. So will it be useful for me to configure data guard.
*2.)* I have configured data guard for testing . My Database is Named as Gard and physical standby database is like stan.
it is working fine.
suppose my main database server get corrupted i have no option to start main database server. Now i left with only standby server how i can start standby server
as database main server.
*3.)* Is there any if main database stop working Standby database start working as main without dba intervention.
Thanks
Umesh
Edited by: Umesh Sharma on Jan 14, 2009 4:17 AM

Hi Umesh,
first of all you should be aware that Dataguard and Replication are two completely different things.
While the database where you replicate to is always up and open in read write mode, a physical standby can either be recovering or open in read only mode.
1.) I think from what you are telling you probably should consider using dataguard with a logical standby database, but be aware that there will be some limitations regarding datatypes.
2.) you can do a failover or a switchover, depending on your database version switchback may not be possible
3.) have a look at dataguard broker, you can use it to automate the failover
Best regards,
PP

Load Balancing and Asynchronous Multimaster Replication

We are planning a new project with 2 server lines for load balancing reasons. We also plan to use sessions being valid for 30 minutes. Each of the 2 server lines has its own Oracle database (identical tables) for the transactions. Within one session the load balancing sends the user always to the same server line.
Independant transactions are executed on both server lines.
Is Asynchronous Multimaster Replication the adequate solution to have the two databases synchronized, what has to be considered (network, memory, cpu) or what else can (has to) be used ?
We want to avoid a RAC solution because of the high costs.

The two options that come to mind here are asynchronous multimaster replication and Oracle Streams.
- What version of the Oracle database will you be using?
- What sort of transaction volume do you anticipate? How much redo generation?
- What sort of connection exists between the two machines?
- Is there an upper limit on the time that can be taken to replicate a transaction from one machine to the other?
Justin
Distributed Database Consulting, Inc.
http://www.ddbcinc.com/askDDBC

Synchronous multimaster replication Backup strategy using RMAN

Hi all,
I am using synchronous multimaster replication. my question is backup strategy has to perofirm in all the master site or only in any of 1 mastersite? what all problems will encounter in RMAn backup in my scenario please help me out with your suggestion
nagaraj.K
[email protected]

You ask: "I want to configure backup strategy using RMAN. any one can help me that"
And the answer is clearly no we can not.
An RMAN backup strategy depends on your SLA (Service Level Agreement) with your customers that you didn't post. What is your down-time window? What is your skill set?
You ask: "How to configure for RMAN Incremental backup?"
Read the docs at http://tahiti.oracle.com for your version (which you didn't mention).
You ask: "What will be backup space and there size ?"
We have no idea. Are you going to create an image copy or a backup set? Read the docs and learn the difference. Are you going to turn on compression? Are you going to back up only changed blocks? We don't know.
You ask: "how to manage growing online archiving files?"
Again we can't help you without knowing your SLA. How many MG/GB are they? What period of time do you need to retain them to meet your SLA? When you figure this out back up to tape those you don't need to keep on the server.
You ask: "how to manage growing data and there disk space?"
This is one we can answer: BUY MORE DISK! That was easy.
You ask: "How we can give good performance our CPU and memory?"
Do you really expect that we can with zero knowledge of your environment, your version, your application, etc. distill into a few short paragraphs the collective wisdom of Cary Millsap, Jonathan Lewis, and the rest of the Oak Table Network membership?
Of course we can not. So what you need to do is buy Cary's book, Jonathan's book, and take their classes.
You ask: "we need keep all archive log in backup files or we need to remove old archive files?"
Remove the old ones after being absolutely sure you have a good backup.
You ask: "where we can take backup tape drive,SAN,disk removable hard disk? which one is better?"
No one solution is better than the other. They are all adequate.

Maximum number of nodes in Multimaster Replication

Has anyone setup a multimaster replication environment on 8.1.7 with more than 11 nodes?
What could be the maximum number of nodes for reliable operation?

Hi,
With Oracle 8.1 you can have upto 9; but I think you can have under one of the replicated DB another replicated DB. tree like tructure.
Best,
EA

Multimaster replication error

Hi,
Well I am a newbie to multimaster replication. I tried to replication few tables and indexes of hr schema to other database but I am receiving the following errors:
ID REQUEST ERRNUM MESSAGE
43 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.EMPLOYEES does not exist or is inv
121 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.DEPARTMENTS does not exist
41 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.DEPARTMENTS does not exist or is i
141 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.DEPARTMENTS does not exist
161 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.EMPLOYEES does not exist or
21 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
22 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
123 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.EMPLOYEES does not exist or
24 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
25 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
26 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
ID REQUEST ERRNUM MESSAGE
27 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
28 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
29 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
30 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
31 CREATE_MASTER_REPOBJECT -942 ORA-00942: table or view does not exist
47 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.JOB_HISTORY does not exist or is i
127 GENERATE_INTERNAL_PKG_SUPPORT -23308 ORA-23308: object HR.JOB_HISTORY does not exist
Kindly, help me regarding this issue.
Thank you for your cooperation.
Kind Regards,
Adnan Hamdus Salam

You will see "replica busy" errors from time to time. The replication protocol is "one at a time" - if a consumer has more than 1 supplier, only 1 supplier may update the consumer at a time - the other suppliers get a busy signal during the update. Note that updates to 1 supplier may be sent to another supplier and then to the consumer if the first supplier gets a busy signal, so all of your updates should get through pretty quickly.
The most important thing is - do your updates get through to all consumers? Do they get through quickly?

Multimaster Replication problem

Hello,
I've configured a multimaster replication environment with two master sites and a master group whith both of them and asynchronous replication. My problem is that in the master def the master group appeared stopped and I donÂ´t know how activate it, and far over, the administrative requests in my replicated databases aren't executed by themselves, in the masterdef I can't depure them but I've to do it explicitily and in the other master site there is no way to get this requests dissapear.
The main problem is that in the secondary master site the replication status is always quiesced whereas in the masterdef the status is normal if there is no objects in the master group and quiesced when I add some objects even though I generate the replication support for all the new objects.
Any help would be very useful because I'm beginning with replication and I've no found any information in the 2 manuals of Oracle about this topic.
Thanks in advance.
Mabel

you can find reason from dba_repcatlog view. I will suggest that you must recreate whole environment and then create your replication freshly. remeber that you replicated table must have primary key coulmn or you have configured suitable conflict resolution method for these tables. Moreover after generating replication support for the objects check dba_repcatlog view. this view should be empty and there is no error located in this view. if there is any error see message from this view and solve it and regenerate you replication support for the object and again check. when you complete all error then resume your replication activity.

In a MultiMaster Replication to not propagate changes on Second DB

Hello,
I've setup a multimaster replication environment with DB1 and DB2.
DB1 replicates all changes on some tables to DB2. But i don't want to DB2 to propagate changes to DB1. So, i didn't create a scheduled link on DB2 to propagate changes to DB1. But deferred transaction queue is getting too large on DB2. How can i settle the DB2 to not to put the changes to deferred transactions?
Regards

Replicating from DB1 to DB2 and allowing local updates on DB2 without propagating them back to DB1 is done with WRITABLE MATERIALIZED VIEWS, not MultiMaster Replication.
Changes made at DB2 are "local" until they are overwritten by changes from DB1.
See http://download.oracle.com/docs/cd/B19306_01/server.102/b14226/repoverview.htm#sthref57
Hemant K Chitale
http://hemantoracledba.blogspot.com

Is schema replicated in multimaster replication scenario?

hellos
I am asking basic question. I have set up a multimaster replication (ids5.1sp2) on database c=fi.. this works well.
however, in the future I know that more objectClasses and attributes will be added to the schema.
My simple question.. is do I have to amend the 99user.ldif files on BOTH ids5.1 servers or can I rely on the schema being replicated.
thanks.

Schema is supposed to replicate in all replication scenario's
The documentation states
"The best way to guarantee schema consistency is to make schema modifications on a single master server, even in the case of a multi-master replication environment.
Schema replication happens automatically. If replication has been configured between a supplier and a consumer, schema replication will happen by default."
However a few people in this forum have complained that this replication doesn't always work - so be sure to check to make sure the schema does in fact replicate.
For more information on schema replication check out this link:
http://docs.sun.com/source/816-5609-10/rep.htm#1063947
It is at the bottom of the page ...

Asynchronous multimaster replication and DB shutdown

Hi
I have a question:
What Happen if in replication environment (Asynch multimaster) master defenition site or other master site shutdown?
Are transaction remained in deffered queue?

Thank you, so I should write in documentation that shutdown Abort is forbidden

Multimaster Replication and VPD in Oracle 10g R2

Hello,
I would like to know if MlitiMaster Replication supports the VPD(Virtual Private Database) Row level or column level access control? if so could some one point me to the right documentation? i searched Metalink and googled but unable to find any info regarding the support of VPD to replication.
Thank you....

Justin,
we have streams and replication in our environment. our databases on window 2003 server and oracle 10gR2. my question is we are planning to implement the VPD on few tables which are already in replication group. We use 'REPADMIN' for replication. will it propagate the data for VPD columns that are in replication environment? how will this impact performance wise? I am trying to find also some Pros and Cons too.
Thank you.

Idenity Sync Windows 1.0 and Multimaster Replication

Does anyone know if IDSW works with Replication? I see in the documentation that it should, but as usual the information is sparse. A couple questions I have:
Do I need to install the SUBC pluggins onto the servers?
What about the changelog ?

David,
The SSL related parameters in the IdSync CLI only apply to the communication between the command line and the configuration directory server.
The communication type to be used during the resync operation is the same as during normal synchronization. This is what you configured with the management console, when you setup your Active Directory and Sun Directory sources.
The -i NEW_USERS option can only be used when resync-ing from Active Directory to Sun Directory server.
This is why you get the error message "Cannot resete DS Passwords when resyncing from DS to Windows".
Bertold

Oracle9i replication transmitted data overhead

Hi
How much replication data overhead is incurred over the network with Synchronous Oracle9i replication over-and-above the data which is being replicated?
I'm sure this will vary depending on the objects being replicated and the data types.
Any feedback on points to further information would be appreciated.
cheers Luie
[email protected]

Assuming that we're talking about Oracle Data Guard, the only thing that gets transferred over the network is the archive logs. If you can determine the rate at which you'll generate redo on the primary, you'll know how much network traffic you'll get.
Justin
Distributed Database Consulting, Inc.
www.ddbcinc.com/askDDBC

Is Multimaster Timesten replication a good option for huge data volumes?

Hi,
There are 3 timesten nodes in our production setup .There will be around 5 million rows in each node initially which will gradually increase to about 10 million.Once our application moves to production, there will be around 50-70 transactions per second in each node , which need to be replicated on to the other node.
Initially we thought of going with Active-Standby-Subscriber replication.However in this case if active and standby node go down,then it will be a site failure case.So is Active-Active (Multimaster replication) configuration a good option ? Will data collision happen when replication happens in both directions?
Thanks in advance.
Nithya

Multi-master replication is rarely a good idea. You will get data collisions unless you rigorously partition the workload. Conflict detection and resolution is not adequate to guarantee consistency over time. Recovery back to a consistent state after a failure is complex and error prone. I'd strongly advise against a multi-master setup, especially for a high volume system.
You seem to be concerned that 2 out of the 3 systems may fail resulting in a site outage. The likelihood of that is small if you have set things up with separate power etc. With the A/S pair based approach you would still have query capability if the two master systems failed. The chances of all 3 systems failing is not that much less than of just 2 failing in reality I would say (depending on the reason for the failure).
Chris

Why Does Replication requires db restart to get transactions through

I am having problems with the deferred tranasactions. There were about 44,000 in the queue before. The dbms job was broken so I ran it and it was able to push half the tranactions from the 44,000 through.
However, only half. I am manually pushing the transactions through using OEM. But they do not seem to go through. Any ideas as to why they do not go through. Any alternatives to get all the deferred tranactions to being pushed?
Oracle 9.2.0.1 on W2k Svr.
The deferred tranaction queue had then incresed in size to 53,000 tranactions to be applied. I ran this query today, and get the following result.
SQL> SELECT SID, r.JOB, LOG_USER, r.THIS_DATE, r.THIS_SEC
2 FROM DBA_JOBS_RUNNING r, DBA_JOBS j
3 WHERE r.JOB = j.JOB;
SID JOB LOG_USER THIS_DATE THIS_SEC
54 8 REPADMIN 15-OCT-06 17:24:52
SQL>
Any ideas as to what to do with this information. I ran the query several times within an hour, but the ame result...any sggestions as to how to apply all the tranactions in teh deferred tranaction queue sucessfully across?
I executed the dbms jobs (which were broken) on all sites so that the DTQ empties. This was sucessfull on two of thethree master sites. However, on one of the master sites, it was only able to push half of the tranactions to one other site. This is what I am trying to resolve.
I try to manually push these transactiosn form OEM, but they do not seem to go through. Is there a way I could find wher the errors lay? I have checked all the alert logs and the trace files but no major errors.
W2K Ad Svr, Oracle 9i Release 2. and three Multimaster sites.
I even ran the following script, it executes, but does not push it through.
declare rc binary_integer; begin rc := sys.dbms_defer_sys.push(destination=>'DESTHOST.DOMAIN', stop_on_error=>FALSE, delay_seconds=>0, parallelism=>0); end;
Once again...in our Oracle 9.2.0.1 multimaster replication environmment on Windows 2000 server system containing three master sites is giving strange errors.
The transactions within defferred transaction queue at one of the sites are not being pushed at the other site. Last month, it was the other way around.
In the trace files, we can see that there are TNS connect error to the other master sites. However as soon as we restart the particular server, the transactions are pushed through.
This environment was working fine before and we were wondering as to what could be causing these errors all of a sudden.
These are only the transactions from the Deferrerd Transaction Queue which are yet to be applied to the other site. The error queue is quite constant and we are not facing much problems there.
In the trace files the errors are:
ORA-12012: error on auto execute of job 78
ORA-23324: error ORA-12535: TNS:operation timed out, while creating deferror entry at "globaldbname" with error -12535
ORA-06512: at "SYS.DBMS_DEFER_SYS", line 1716
ORA-06512: at "SYS.DBMS_DEFER_SYS", line 1804
ORA-06512: at line 1
*** 2006-11-07 22:05:42.000

Did you fix your problem?
If Yes please help me bescause I am in the same situation

Multimaster replication survival & overhead

Similar Messages

Maybe you are looking for