Partial transaction for multi-master asynchronous replication

I have a fundamental question about multi-master asynchronous replication.
Lets consider a situation where we have 2 servers participating in a multimaster asynchronous mode of replication.
3 tables are part of an Oracle transaction. Now if I mark one of these tables for replication and the other 2 are not a part of any of the replication groups.
Say as a part of Oracle transaction one record is inserted into each of the 3 tables.
Now if I start replicating, will the change made in the table marked for replication be replicated on to the other server. Since the change made to the other 2 tables are not propogated by the deferred queue.
Please reply.
null

MR.Bradd piontek is very much correct.If the tables involved are interdependent you have to place them in a group and all of them should exist at all sights in a multi master replication.
If the data is updated(pushed) from a snapshot to a table at a master site it may get updated if it is not a child table in a relationship.
But in multi master replication environment even this is not possible.

Similar Messages

Multi Master Site replication with Snapshot features

Dear,
I am implementing a Multi Master site replication. I want to replicate Table A in MASTER SITE A to Table A in MASTER SITE B.
However, I don't want all the data from MASTER SITE A to be propagated to MASTER SITE B.I may want data that belongs to a certain criteria only to be replicated over.
I know I can achieve this in a snapshot site replication. But can this be done in a Multi Master Site replication too ?

Hai,
As far under my observation, the tables that is marked for a replication from the MASTER SITE to Child Site is an exact replica of data. Your case could be achieved only thru SNAPSHOT VIEWS/GROUPS
- [email protected]

IDM in multi-master LDAP Replication

Hi,
We got two functional SUN Java Directory Server in multi-master replication setup. Both server have their own IDM's.
When I change password/uid from any IDM , straightaway changes get done on both LDAP servers and I can see changes on another IDM.
Problem is when I create new user from IDM of one server, user doesn't show up in second server IDM unless I run manually Accounts-->Load from resource.
Even full reconciliation doesn't pickup the new user on that IDM. What need to be done so IDM picks new users straight away in multi-master setup.
Thanks,
Farhan
Edited by: rozzx on May 5, 2009 11:32 PM
Edited by: rozzx on May 5, 2009 11:34 PM

Any help guys? Whey IDM is not getting update when I add/delete new user in Directory Server. I have to do Load from Resource to get new entries everytime.
And If I delete any user from LDAP, it still stays in IDM.

Can the ApplicationModule support global Transaction for multi db instance?

purpose:
I want to update data in two db instance .
pre-conditions:
I create two application module , in the two application modules ,one is root application module. one application module connect one db instance, the other application module connect to the other db instance.The two application modules include entity objects and view objects which map the tables in the two db instance.
In the jsp page , I bind two view objects , one view is owned by root application module ,the other view object is owned by child application module and the two view objects map the two tables which are in different db instance .the view objects binding style in the jsp page is input-form.
The result:
while the input-form is submitted ,the error occurs.The global transaction isn't committed, why?

A nested application module instance uses the transaction/connection of its parent.
The simplest way to accomplish what you're trying to do is to use Oracle database DBLinks so that the two-phase commit is handled at the database level.
ADFBC does not support two-phase commit at the application level between two separate root application module's transactions.

Multi-master Replication and Metadirectory 5.0

The Metadirectory 5.0 documentation states that it cannot work with a directory server configured for multi-master replication. We need to use Metadirectory since we are integrating the Directory Server with other systems. Does this mean that we'll be forced away from MMR configuration? What are some of the alternatives? Does iPlanet have any plans for supporting MMR in future versions of Metadirectory?

I think you can enable the retro changelog on a consumer replica. I'm pretty sure that works.
You might be able to enable it on a hub. You also might get it to work on a master, but the changelog on the master will contain some replication housekeeping changes that may confuse Meta Directory. I'm not sure what they are but they are obvious if you look at the changelog. If you can figure out some way to screen those changes out from Meta Directory, you might be able to use it.

Custom Transaction for RSDMD

Hello Friends,
I am looking to create a custom transaction for RSDMD (Master Data Maintenance) for a specific MDInfoObject ( say for example ZIO_TEST1).
I did try following:
1. Used SHD0 - created a Variant.
    I did see the message Variant "TVAR_RSDMD" saved.
2. Used SE93 to create ZVAR1
      Selected fourth radio button-
         Transaction With Variant
3. In the next screen -
    a. entered transaction text - some desc.
    b. for transaction entered - RSDMD
    c. When I try to enter transaction variant - I get an error message "No transaction variant exists."
      I cannot see the drop-down list also.
Please help!! help!!.
Thank you,
Pramod.

Hi,
Transaction variant?
http://help.sap.com/saphelp_nw04/helpdata/en/67/232037ebf1cc09e10000009b38f889/content.htm
Transaction Variants
http://www.erpgenie.com/abap/tvar.htm
Regards,
San!

Multi-master replication questions for iPlanet 5.0, gurus out there?

hi:
I'm using iPlanet Dir Server 5.0 and I note that many gurus out there has
been able
to get this to work, that's good, but I have yet to. I have several
questions, maybe
someone can spend a few minutes and save me hours...
I have a suffix called dc=calient,dc=net. I followed the suggestions in
the
iPlanet install guide and created 2 directory servers
a) suffix o=NetscapeRoot, at some arbitrary port, 4601
b) suffix dc=calient,dc=net, at the usual port 389.
All my searches/create/delete work fine. However, when I try to replicate
with multi-master between 2 machines, I keep getting into problems.
Here's one set of questions...
Q1: do people out there really split their tree from the o=NetscapeRoot
tree?
Q2: The admin guide says the the unit of replication is a database, and
that each replication can only have 1 suffix. Is this true? Can
a replicated db have more than 1 suffix?
Q3: If I also want to replicate the o=NetscapeRoot tree, I have to set
up yet 2 more replication agreements. Isn't this more work? If
I just lump the 2 suffixes together, wouldn't it be easier? But would
it work?
Q4: I followed the instructions to enable replicas on the masters.
But then I tried to create this cn=Replication Manager, cn=config
object.
But what is the object class of this entry? An iPlanet user has uid
as its RDN... I tried a person object class, and I added a password.
But then I keep getting error code 32, object not found in the error
log. What gives? such as
WARNING: 'get_entry' can't find entry 'cn=replication
manager,cn=config', err 32
Q5: Also, are there any access control issues with this cn=Replication
Manager,
cn=config object? By this I mean, I cannot seem to see this object
using
ldapsearch, I can only see cn=SNMP, cn=config. Also, do I have
to give all access via aci to my suffix dc=calient,dc=net? Also,
given the fact that my o=NetscapeRoot tree is at a different port (say
4601),
not 389, could this be an issue?
Q6: when replication fails, should the Dir Server still come up? Mine does
not anymore
which is strange. I keep getting things like this in my log file
[08/Nov/2001:21:49:13 -0800] NSMMReplicationPlugin - Could not send consumer
mufasa.chromisys.com:389 the bind request
[08/Nov/2001:21:49:13 -0800] NSMMReplicationPlugin - Failed to connect to
replication consumer mufasa.chromisys.com:389
But why shouldn't the dir server itself come up even if replication
fails?
steve

Hi Steve,
First, please read the 'Deployment Guide'. I think that is easier to
understand when you want to setup multi-master replication. The
'Administrator's Guide' gives you step-by-step instructions, but it may
not help you to understand how to design your directory services.
Stephen Tsun wrote:
I have a suffix called dc=calient,dc=net. I followed the suggestions in
the
iPlanet install guide and created 2 directory servers
a) suffix o=NetscapeRoot, at some arbitrary port, 4601
b) suffix dc=calient,dc=net, at the usual port 389.
All my searches/create/delete work fine. However, when I try to replicate
with multi-master between 2 machines, I keep getting into problems.I don't understand something: which backend do you want to replicate?
The one holding 'o=NetscapeRoot' or the one holding 'dc=calient,dc=net'?
Do you want to setup replication between these two instances of the
directory server (i.e. between port 4601 and 389 in your example)?
Q1: do people out there really split their tree from the o=NetscapeRoot
tree?If you have multiple directory servers installed in your environment, it
is probably worth dedicating (at least) one directory server for the
o=netscaperoot tree.
Q2: The admin guide says the the unit of replication is a database, and
that each replication can only have 1 suffix. Is this true? Can
a replicated db have more than 1 suffix?Well, it is normal, since in iDS 5.x you have 1 suffix per database.
You can, however, replicate multiple databases.
Q3: If I also want to replicate the o=NetscapeRoot tree, I have to set
up yet 2 more replication agreements. Isn't this more work? If
I just lump the 2 suffixes together, wouldn't it be easier? But would
it work?You can't lump the 2 suffixes together, because each backend has 1
suffix associated with.
Q4: I followed the instructions to enable replicas on the masters.
But then I tried to create this cn=Replication Manager, cn=config
object.
But what is the object class of this entry?Usually, it is organizationalperson or inetorgperson. In most of the
cases you want an objectclass which can have userPassword attribute.
An iPlanet user has uid
as its RDN... I tried a person object class, and I added a password.
But then I keep getting error code 32, object not found in the error
log. What gives? such asYou must have misconfigured something. Or perhaps, it is not
cn=replication manager, cn=config, but 'uid=replication manager,cn=config'
Q5: Also, are there any access control issues with this cn=Replication
Manager,
cn=config object? By this I mean, I cannot seem to see this object
using
ldapsearch, I can only see cn=SNMP, cn=config.The configuration tree is protected by ACIs, so you can not see them
using anonymous BINDs. Try binding as 'directory manager' and you will
find your entry.
Also, do I have
to give all access via aci to my suffix dc=calient,dc=net?For what purpose? For replication, it is enough to set user DN in the
replication agreement and this user can update the replicated backend.
Q6: when replication fails, should the Dir Server still come up?Yes.
Bertold

Multi Master Replication - Only works for some tables?

I have a multi master replication between a 9i and an 816 database.
All the tables are in the same tablespace and have the same owner. The replication user has full privs on all the tables.
When setting up the replication some tables create properly, but others fail with a "table not found" error.
Ideas anyone ?
Andrew

You said that you have a 9i replicated with a 816.
I try the same thing but with two 9i enterprise version databases, free downloaded from www.oracle.com.
when i ran
exec dbms_repcat.add_master_database(gname=>'groupname', master=>'replica_link')
this error appears
ERROR at line 1:
ORA-23375: feature is incompatible with database version at replica_link
ORA-06512: at "SYS.DBMS_SYS_ERROR", line 86
ORA-06512: at "SYS.DBMS_REPCAT_MAS", line 2159
ORA-06512: at "SYS.DBMS_REPCAT", line 146
ORA-06512: at line 1
please help me if u have any idea.

DS 5.1 "half" multi-master replication

I realize the situation described below is not the best method, but those decisions were out of my hands.
I have two servers in two datacenters (serverA and serverB). Both servers are set as multi-masters within the console. However, due to replication over WAN restrictions, there is only one replication agreement from serverA to serverB (no agreement from serverB to serverA). All writes are directed to serverA and serverB is not used by any clients for queries. ServerB is intended as a hot failover should we need it (again, don't ask, not my decision).
We have been experiencing the second host (serverB) randomly crashing. There are no indications within the logs (of course minimal error logging is on).
My bet is the crashing is due to the "half" nature of this multi-master set up. Has anyone else seen or can confirm this? Again, yes I know this is not the ideal way to go...sometimes you just have to do what you're told.
Thanks

One thing to check would be the replication logs of
the master1 server to see if it is having
difficulties prior to the crash of master2.
Will have to check the logs. The logging level is usually turned fairly low due to the number of changes and utilization on the server (master1 not master2).
Is master2 used for anything else (application
queries or other read-transactions)? The thinking
being that system utilization or excessive
connections could be causing issues ...Master2 is not used for anything. No queries other than regular monitoring queries ("Can you hear me?") We have had one of our dedicated consumers crash similar to this, but only once not the number of times (up to about 6+ over the past month) the secondary master has crashed.
The server is a Sun 480R running Solaris 8. The servers in the farm are supposed to be near-identical. It has come to the point where I am debating whether it is a replication problem (as I noted in my original post) or the system just needs to be rebuilt.

Multi-master replication - indefinite referrals

Hi,
I am trying to set two master ldap in multi master mode on college campus:
LDAP1 and LDAP2.
I get this in errorlog of LDAP1:
Configuration warning This server will be referring client updates for replica o=comms-config indefinitely
WARNING<20805> - Backend Database - .. - search is not indexed
I guess the warning says the search is not indexed, but still that's not the mail problem...
in other server logs on LDAP2, I get:
ldap://LDAP2:389; Partial results and referral
received|#]
...0400|SEVERE|sun-appserver|javax.enterpr
ise.system.container.web|_ThreadID=24|SomeJSP:Servlet.service() for
servlet jsp threw exception java.lang.Exception: The Error has occured due to
:netscape.ldap.LDAPReferralException: referral (9); Referral:
LDAP2; Partial results and referral received at...
dse.ldif from LDAP1:
nsslapd-referral: ...
numsubordinates: 1
nsslapd-state: referral on update
from LDAP2:
nsslapd-referral:...
nsslapd-state: backend
Should LDAP1 and LDAP2 be both set to "process write and read requests" ??
Please advise

The state of LDAP1 is probably the result of the way you've done the initialization of both servers and replication.
What to do in this situation is precisely explained in the Administration Guide, Replication chapter.
Regards,
Ludovic

Latency at the apply side in multi master replication

Hi Gurus,
We are facing some multi master replication latency always at one side . Ex: Site1 -> Site 3 . Transactions from site 1 always gets applied on Site 3 with some latency though the latency doesnt happen while applying from Site 1 - Site 2 for the same transaction originating at site 1 in 3 site(3 way) multi master replication environment.
We have investigated into it and now looking to check the system replication tables which would be involved at the apply side (ex:- Site 3).
Could someone pls let me know all the system replication tables that would be involved at the apply side which can impact the latency .I know few of them but dont know the all of them.
System.def$_AQERROR
System.def$_ERROR
System.def$_ORIGIN
Thanks

I would say that 50, 50 and 75 are a Very Large number of Job Queue processes. Do you really have that many jobs that need to run concurrently ?
Since Advanced Replication Queues are maintained in only a small set of tables you might end up having "buffer busy waits" or "read by other session" waits or latch waits.
BTW, what other factors did you eliminate before deciding to look at the Replication tables ?
See the documentation on monitoring performance in replication :
http://download.oracle.com/docs/cd/B10501_01/server.920/a96568/rarmonit.htm#35535
If you want to look at the "tables" start with the Replication Data Dictionary Reference at
http://download.oracle.com/docs/cd/B10501_01/server.920/a96568/rarpart4.htm#435986
and then drill down through the View definitions to the underlying base tables.

Problem in multi master replication creation using DBA Studio -- replication

Hi,
I am trying to create multi master replication using DBA studio but facing following problem at the time of master group creation.
ORA-04052 error occured when looking up remote object SYS.SYS@CBOLDATA
ORA-00604 error occured at recursive SQL level 2
ORA-01017 invalid username/password, logon denied
ORA-02063 preceding line from CBOLDATA
If you want to know how I am trying the whole thing then here is the way I am doing comfiguration.
First I have created master site which is created successfully.
Here I have used two database named UPP817 & CBOLDATA
Step of master site creation is followed like this.
1. Master Site addition (Added both UPP817 & CBOLDATA using SYSTEM username)
2. Default User (No change is done, Default schema and password taken)
3. Master Site Schema (Added SCOTT/tiger)
4. Scheduled Link (No change is done, Default values taken)
5. Purge (No change is done, Default values taken)
6. Finished successfully
Master group creation there is three option available
1. General (Only value of Name is given)
2. Object (No object added)
3. Master Sites (It takes UPP817 by default as a master defination site then as a Master Site I have added CBOLDATA)
when I pressed on Create It gives me the error listed above.
I really appreceate your help
Mukesh

Create public database link at master site database for CBOLDATA.
Also see that you are using SNMP protocol and Oracle Pipe at both
databases. This will be usefull while making replication from remote
place.
I hope this will work.
regards
Avinash Jadhav

Multi master replication between 5.2 and 6.3.1

I have a setup in which I have a master running version 5.2 and about 15 consumers ( slaves) all of which have been upgraded to 6.3.1 . I now want to create a multi master topology by promoting one of these consumers to be a master and still keep the 5.2 in use as we have a bunch of other applications that depend on the 5.2 instance. Our master has two suffixes. The master server is also the CA cert authority for all the consumers . After reading the docs I narrowed down the procedure to be
1. Promote one of the 6.3.1 consumers to hub and then to master using the dsconf promote-repl commands. The problem here is that I am not sure how I can create a single consumer that can slave both the suffixes. We currently have them being slaved to different consumers.
Also do I need to stop the existing replication between the 5.2 master and the would be 6.3.1 master to promote to hub and master.
2. Set the replication manager manually or using dsconf set-server-prop on the new 6.3.1 master .
3. Create a new replication agreement from 5.2 to 6.3.1 master without initializing. (using java console)
4. Create new replication agreement from 6.3.1 to 5.2 (using command line)
5. Create new repl agreements between the new 6.3.1 master and all the other consumers. For this do I need to first disable all the agreements between 5.2 and 6.3 or can I create new agreements without disabling the old ones?
6. Initialize 6.3.1 from the 5.2 master.
My biggest concern at this point is surrounding the ssl certs and the existing trusts the consumers have with the 5.2 master. Currently my 5.2 server acts as the CA authority for our certificate management with the ldap slaves. How can I migrate this functionality to the new server and also will this affect how the slaves communicate to the new master server ?
Thanks in advance.

Thanks Marco and Chris for the replies.
I was able to get around the message by first manually initialzing the new slave using an ldif of the ou from the master , using dscc to change the default replication manager account to connect and finally editing the dse.ldif to enter the correct crypt hash for the new repl manager password. After these steps I was able to successfully set up replication to the second ou and also promote it to hub and master ( I had to repeat the steps after promotion of the slave to master as somehow it reset replication manager settings when I did that).
So right now, I have a 5.2 master with two ou's replicating to about 15 consumers.
I promoted one of these to be a second master (from consumer to hub to master). Replication is setup from 5.2 to 6.3 master but not the other way round.
I am a little bit nervous setting up replication the other way round as this is our production environment and do want to end up blowing up my production instance. The steps I plan on taking are , from the new master server
1. dsconf create-repl-agmt -p 389 dc=xxxxx,dc=com <5.2-master>:389
2. dsconf set-repl-agmt-prop -p 389 dc=xxxxx,dc=com <5.2-master>:389 auth-pwd-file:<passwd_file.txt>
I am assuming I can do all of this while the instances are up. Also in the above, does create-repl-agmt just create the agreement or does it also initalize the consumer with the data ? I want to ensure I do not initialize my 5.2 master with my 6.3 data.
Thanks again

Procedure to create Multi Master Replication

Hi,
can anybody help on this. How to do Multi Master replication from scarch . I used to do single master . this is the first time iam trying to do Multi-Master Replication . I am looking for doc, iam unable to find . Is ther anybody can tell how to do Multi-Master little in detail. I installed my Software and i loaded my company schema and the data on one server. I have another server ready to install software and load data.

with regards to the post "RTFM"..
in my experience the sunone/iplanet manuals are very badly written and in some cases just plain wrong..
sure start with reading the manuals, then make sure you also read the release notes and then you can still be confused in some cases....
the section on mmr in the manual has a number of mistakes last time i looked..
one reason why i think community sites, and generally available "support" are such prime requisites in choosing software...
just my 2 cents

Using attribute uniqueness with multi-master replication?

Hi,
I'm trying to use attribute uniqueness in a iDS 5.1 multi-master replication env. I have created a plug-in instance for the attribute (memberID) on each directory instance (same installation on NT) and tested (if I try to create a duplicate value under the same instance I get a constraint error as expected). However if I create a entry under one instance and then create a second entry (different DN) with the same attribute value on the second instance, the entry is written with no complaints? If I create the entries with an identical DN, then the directory automatically adds nsuniqueID to the RDN of the second entry to maintain DN uniqueness but it doesn't seem to mind about duplicate values within the entry despite the plug-in?
BTW I've tested MMR and it is working and I'm using a subtree to enforce uniqueness.
Regards
Simon

Attribute uniqueness plugin only ensure uniqueness on a single master before the entry is added. It doesn't check replicated operation since they have already been accepted and a positive result was returned to the client. So in a multiMastered environment, it is still possible to add 2 identical attributes, if like you're saying you're adding the entries at the same time on both master servers.
We're working on a solution to have Attribute Uniqueness working in a multiMastered environment. But we're worried about its impact on performances we may get with it.
Regards,
Ludovic.

Partial transaction for multi-master asynchronous replication

Similar Messages

Maybe you are looking for