What is transaction replication used? I seemed to create transaction replication following this tutorial:
http://www.sql-server-performance.com/2010/transactional-replication-2008-r2/
And I know when I change some objects i.e any DML or DDL statement, those changes will be reflected to the other server where I did replication. But it's not clear to me why should we use transaction replication. Does SQL server automatically start using the 2nd server where replication was done when the main instance fails? or do we have to manually restore the database from the server where replication was done in case of failure of 1st instance?
Thanks in advance :)
You can use transactional replication to maintain a warm standby SQL server. Transactional replication replicates the data on one server (the publisher) to another server (the subscriber) with less latency than log shipping.
You can implement transactional replication at the database object level such as the table level. Therefore, Microsoft recommends that you use transactional replication when you have less data to protect, and you must have a fast data recovery plan.
This solution is vulnerable to the failure of the publisher and the subscriber at the same time. In such a scenario, you cannot protect your data. In all other scenarios such as the failure of a distributor or a subscriber, it is best to resynchronize the data in the subscriber with the data in the publisher.
You should use transactional replication to maintain a warm standby SQL server only when you do not implement schema changes or you do not implement other changes to your database such as security changes that replication does not support.
Note Replication is not designed for the maintenance of warm standby servers. With replication, you can use replicated data at the subscriber to generate reports. You can also use replication for other general uses without having to perform processing on your relatively busy publisher.
Disadvantages
Schema changes or security changes that are performed at the
publisher after establishing replication will not be available at
the subscriber. The distributor in transactional replication uses an
Open Database Connectivity (ODBC) connection or an OLE Database
(OLEDB) connection to distribute data. However, log shipping uses
the RESTORE TRANSACTION low-level Transact-SQL statement to
distribute the transaction logs. A RESTORE TRANSACTION statement is
much faster than an ODBC connection or an OLEDB connection.
Typically, switching servers erases replication configurations.
Therefore, you have to configure replication two times:
a. When you switch to the subscriber.
b. When you switch back to the publisher.
If a disaster occurs, you must manually switch servers by
redirecting all the applications to the subscriber.
Read more here http://sqlserverdatarecovery.com/transactional_replication.html
Related
What are the main differences between sql-server replication and availability groups in an Azure Iaas data solution? (sql server instances on a VM)
Are these the same concept?
SQL Server Replication copies data and replays changes using SQL Commands to replicas. Sometimes this is called "Logical Replication". Replication is typically not used for High-Availibility or Disaster Recovery, but can be used to create a readable replica of a production database or to copy selected data from a main database into one or more subscribers.
AlwaysOn Availibility Groups copies and applies the log records to syncronize databases. The changes are applied using the same transaction log redo process that's used in backup/restore and database recovery. Sometimes this is called "Physical Replication", as the database files on every replica are identical. AlwaysOn Availibility Groups can be used for High-Availitbiiy, Disaster Recovery, and to create Readable Replicas of a production database.
Is there anyway to identify by how much time we can get a deadlock when we enable Snapshot replication for our subscriber, we have a subscriber considered as Azure SQL Database, and we have 8 databases one of them consists of 200GB in size. Is there any way to identify the downtime for genereating snapshots?
When the initial snapshot is being generated while configuring SQL Server
Transactional Replication, does anyone know if the snapshot agent places locks on the tables (articles) you have selected? I have some tables that contain 2+ millions rows and wanted to know if SQL Server actually places table locks to prevent updates while the publishing database is online. If locks are placed, then I want to run the initial snapshot during off peak hours in production.
Thanks!
In Transactional replication or any other type of replication the starting point is a snapshot of the database. The initial step of creating the snapshot is exactly the same in any type of the replication.
SQL Server does not obtain any kind of locks at all when creating a snapshot, it literally is a snapshot of the database at a certain point in time and creating snapshot does not interfere with any transactions. Uncommitted transactions are rolled back in the snapshot once it is created.
To read more about how database snapshot works read this article from MSDN How Database Snapshots Work
If you're running on an edition of SQL Server that supports database snapshots (as in create database [foo]... as snapshot of [bar]), then you can optionally use those as the basis of the snapshot. Check the #sync_method parameter of sp_addpublication. The caveat is that you still probably want to do it during a non-busy time of the day because of how database snapshots work (i.e. copy-on-write will slow down any write activity), but you won't be contending on locks.
Starting SQL Server 2005, the default #sync_method value for sp_addpublication is "concurrent", which means the tables are not locked during snaphsot agent run. Note this is not entirely true - the snapshot agent places schema locks on the tables, but the duration of that lock is mere seconds at most.
So if you set #sync_method = "concurrent", then no, updates, in theory, will not be blocked. If #sync_method = "native" (default in SQL Server 2000) or "character", then yes, updates will be blocked.
I have a windows server 2012 with SQL Server 2012 configured with Merge replication with three subscribers.
I want to take a full back up the merge replication in case of total sql server crush. I am not sure how to take it or which databases to take a back up? for how long? so that if some thing happens then I will just restore all the configuration back.
any idea?
thanks
You should take backups of these databases
The publication database at the Publisher
The distribution database at the Distributor
The subscription database at each Subscriber
The master and msdb system databases at the Publisher, Distributor and all Subscribers. These databases should be backed up
at the same time as each other and the relevant replication database.
For example, back up the master and msdb databases at the Publisher
at the same time you back up the publication database. If the
publication database is restored, ensure that the master and msdb
database are consistent with the publication database in terms of
replication configuration and settings.
For more information, refer msdn
To add to what ughai posted, all replication components should be scripted out as a part of a disaster recovery plan. They should also be rescripted if any publication and/or subscription property changes are made.
This is covered in the section Script the replication topology in Best Practicies for Replication Administration and in Scripting Replication.
It does seem like a need of a proper disaster recovery plan, from what you describe. You can create your own or maybe use a third party vendor in order to do that. We worked with cloudendure back at the time, I'm not sure as for their costs right now, maybe it's worth a look. What happens is that you've got a replica being formed and updated in 1 sec time intervals, replication of the data goes right onto the target location (the cloud), and we're aiming for near zero downtime so for whenever that happens, your replica will be loaded automatically onto your site.
SQL Server 2008R2 Transactional Replication - Move SubscriberDB - Push Subscription
I have a requirement to move a subscriber database to a new server outside of a system outage i.e. I cannot stop new transaction from loading into the publisher database.
So far I have attempted to stop the distribution agent and let all un-replicated commands replicate to the subscriber database at Server1. Then backup and restore the subscriber database to Server2. I have then created a new subscription from Server2 to the existing publication.
This works but only transactions created from that point forward are replicated to the Server2 subscriber database. I also require all the old transactions built up at the distributor database destined to only go to Server1.
Is there a replication command available to update the destination of existing transactions at the distributer to the new subscriber.subscriber_DB?
There is 1 publication with multiple articles. The publication is currently only subscribed to by one database at Server1.
You're working too hard. Create a subscription on the new server as though you didn't have an existing subscription at all. There are many ways to do this; pick your favorite. I see in your response to another answer that snapshots are a no-go. I have had a lot of success using the "initialize from backup" option (here, "backup" refers to a backup of the publisher). Once the subscriber is in sync after initialization, you're done. You're now clear to "migrate" to the new server.
Due to the distribution agent being specific to each subscribing server, you can't really have a new subscriber take over where an existing subscriber left off. The only way this would work is if you backup/restore to Server2, temporarily stopped transactions on the publisher, backup/restore log to Server2, create subscriptions with "replication support only", and then re-enable transactions.
Can you not just add subscriptions to the same publication for the new subscriber and kick off the snapshot agent? That will independently sync Server2 while Server1 continues to stay in sync. You should then be able to switch to Server2 seamlessly and remove the subscriptions to Server1.
One note of warning though. I would check the sync_method in syspublications table (or run sp_helppublications) of your publication database to make sure snapshots won't lock the tables in your publisher (by default with SQL Server 2005+, this shouldn't be an issue). Otherwise, what I just suggested will cause blocking on the publisher until the snapshot finishes.
Also, this is a good reference to understanding how to do snapshots one at a time if you setup replication through the GUI using default settings. This will be useful for distributing the snapshots over a larger period of time to reduce publisher server load if you have a lot of data to replicate (gigabytes rather than megabytes).
http://www.replicationanswers.com/TransactionalOptimisation.asp