Recently I received a requirement from application team to configure one way replication although it sound weird but yes this is possible and Microsoft provides some parameters to tweak normal behavior of Merge Replication.
Before we proceed further let’s discuss about normal behavior of Merge Replication and when to use Merge Replication.
Merge replication is one of the modes available in the Microsoft SQL Server for distributing data to various servers from a primary server. Merge replication is one of three types of replication, along with snapshot replication and transactional replication. Which type is used depends on the database’s needs, how frequently changes are made to it and the SQL Server version being employed.
Merge replication is the most complex type of replication because it allows both publisher and subscriber to independently make changes to the database. In this scenario, it is debatable whether the publisher is strictly the primary server, because other servers can also make changes to the data. At any rate, the changes are then synchronized by merge agents that sit on both servers, as well as by a predetermined conflict resolution mechanism in case of clashing data changes. Such clashes may arise because merge replication does not require a real-time network connection between the publisher and the subscriber, which raises the very real possibility of one server changing data, and another server later changing the very same data to a different value.
Here is an article which explains about conflict detection and resolution.
Merge replication is commonly used by laptop and other mobile users who cannot be constantly connected to the publisher, but still need to carry around a copy of the database that they can make changes to.
Are there any disadvantages for Merge Replication?
- It takes lot of time to replicate and synchronize both ends.
- There is low consistency as lot of parties has to be synchronized.
- There can be conflicts while merge replication if the same rows are affected in more than once subscriber and publisher. There is conflict resolution in place but that adds complication.
Why there is a need for one way merge are there any advantages?
Consider a scenario where application needs lesser number of jobs less administration of jobs, distribution database and want to get rid of Publisher to Distributor latency and specifically for handheld\mobile devices\Read only data stores who cannot be constantly connected to the publisher, but still need to carry around a copy of the database.
It is also quite possible that you need this kind of setup as there are some constraint adding primary keys in your database\application and still you need some specific set of data proliferation to other sites. But you have t be careful on this as this will add additional RowGuid columns in table, and that you have to take care on front end.
From the advantages point of view I only see that with this One Way Merge you can get rid of Log Reader Agent, you will not see issues like log reader agent is stuck while scanning thousands of VLF (virtual Log files) when log grows too much in certain scenario and you don’t have to bother much about conflict resolution.
How to configure One Way Merge Replication?
There are 2 ways to achieve this kind of setup where we want data to be pushed from Publisher to Subscriber.
The First way is at the article level where we can decide that a merge table article is download-only with an option whether you want to make changes at the subscriber but they will not be uploaded to the Publisher and another is changes are not allowed at the subscriber at all. This is achieved by using below property.
When adding an article, there is an option to define the “subscriber_upload_options“:
sp_addmergearticle @subscriber_upload_options= subscriber_upload_options
This defines restrictions on updates made at a Subscriber (with a client subscription). The parameter “subscriber_upload_options” is a tinyint, and can have one of the following values.
|0||No restrictions. Changes made at the Subscriber are uploaded to the Publisher.|
|1||Changes are allowed at the Subscriber, but they are not uploaded to the Publisher.|
|2||Changes are not allowed at the Subscriber.|
Hence to achieve this functionality we have an option to choose 1 (Changes are allowed at the Subscriber, but they are not uploaded to the Publisher) or 2 (Changes are not allowed at the Subscriber).
@subscriber_upload_options =1 is defined as “download only, but allow subscriber changes”. In this option there will be no triggers at the subscriber so there will be no firing of triggers to unnecessarily log metadata at the subscriber, which makes both subscriber data changes and the subsequent synchronization significantly faster.
@subscriber_upload_options =2 disallows all subscriber changes. In this case there is a special trigger – MSmerge_downloadonly* which will rollback any attempt to change subscriber data.
Msg 20063, Level 16, State 1, Line 1
Table into which you are trying to insert, update, or delete data has been marked as read-only. Only the merge process can perform these operations.
Once you are done with the configuration of Publication add subscriber where you want to download these articles and subsequent transactions with subscription type property as “Client”.
This is how you can achieve one way merge replication by changing the property of published articles sp_addmergearticle @subscriber_upload_options= subscriber_upload_options
There is another way by which you change the normal behavior of merge replication and force the merge replication into Unidirectional Merge. This can be achieved by changing normal behaviour of Merge Agent using property “EXCHANGETYPE”
The value of -EXCHANGETYPE determines the direction of merge replication changes. This can be done by manually editing the Merge agent job step by adding -EXCHANGETYPE parameter with value 2
- UploadOnly (1): Only changes originating at the Subscriber are merged with the Publisher. The Publisher’s changes stay in the Publisher. Use a value of 1 in your agent properties.
- DownloadOnly (2): Only changes originating at the Publisher are merged with the Subscriber. Use a value of 2 in your agent properties.
- Bi-Directional (3 – Default): Changes originating at the Publisher and Subscriber are merged. Use a value of 3 in your agent properties.
|UPLOAD||1||Only merge Subscriber changes with the Publisher.|
|DOWNLOAD||2||Only merge Publisher changes with the Subscriber.|
|BIDIRECTIONAL||3||Merge all changes between the Publisher and Subscriber (default).|
As soon as we configure with parameter with value of 2 it means that changes to a replicated article at the subscriber are not prohibited, are recorded in the merge metadata tables via merge triggers, and are subsequently filtered out when the merge agent synchronizes. This means there may be a huge amount of metadata unnecessarily recorded, slowing down data changes and synchronization.
Conclusion: One Way or Unidirectional replication can be achieved easily in Merge Replication with minor tweaking in SQL 2005 and later version even at very granular level.
sp_addmergearticle @subscriber_upload_options =1 parameter defines restrictions on updates made at a subscriber. The parameter value of 1 is described as download only, but allow subscriber changes and seems equivalent to the -EXCHANGETYPE = 2 setting mentioned previously, but in the SQL Server 2005 case there are no triggers at all on the subscriber table. Another distinction is that this setting is made at the more granular article level rather than set for the entire publication. This means that although the -EXCHANGETYPE and sp_addmergearticle methods are logically equivalent, the implementation has become much more sophisticated in SQL Server 2005 and later versions. Triggers that unnecessarily log metadata at the subscriber are no longer fired; therefore both subscriber data changes and the subsequent synchronization are significantly faster.
@subscriber_upload_options =2 disallows all subscriber changes. In this case there is a special trigger MSmerge_downloadonly* which will rollback any attempt to change subscriber data.
However you have to be very sure while using this option as this option is depreciated feature list and about Re-initialization of subscription with this kind of setup.
Deprecated Features in SQL Server Replication
I would highly appreciate feedback and comments on this article and would love to know more about any advantages you are getting from this kind of setup if you are already using this Unidirectional Merge.