SQL Server Performance

Tran & snapshot replication on A/P Cluster

Discussion in 'SQL Server Clustering' started by deepakontheweb, Mar 8, 2005.

  1. deepakontheweb New Member

    Hi all:

    I have 2 node active passive clusture, over it i need to establish snapshot replication for 3 db's and transactional replication for 4 databases.
    Actually we have requirments that need real time updated databases for reporting applications, thus i can't use logshipping..

    Please give me ideas, what all precautions i should keep in mind before implementing on production databases. i have tested both kind of replication of test environment but publisher was not a clusterd server.

    i will be using active server as publisher and for distributor too, reporting server will be using pull replication for changes.

    all three server have identical models and 4 number of cpu's init on windows 2003 Ent. Edition and sql 2000 ent edition with sp3a under same domain.

    --Thanks in advance..






    Deepak Kumar
  2. deepakontheweb New Member

    all three servers means
    Active
    Passive and
    Reporting.


    I have heard that implementing replication on clusterd server's will be hard to manage specially in my case because i have almost 200 transactions per seconds.. and daily all database increment is almost 300 to 400 MB.

    another i take full sql backup on daily basis at night, do 1/4 database indexdefrag daily early morning that fillup log file to 15 gb.. plus do transaction log backup after every 10 minutes.

    oooops..


    Deepak Kumar
  3. Argyle New Member

    That you are using a cluster will not be much different than a single server, if at all. An A/P cluster is just like a single server but with a passive node that can take over the services. You will have same performance and design questions to consider for a clustered setup as you have for a non clustered setup. So I don't know where you have heard that it would be more difficult to manage just because it's a cluster and you have many transactions.

    I would not focus on that you have a cluster in the environment. I would focus on how to set up your replication scenario in an optimal way as you would have done with a single server and then just apply the same concept on your clustered server.
  4. deepakontheweb New Member

    Thanks Argyle, I have implemented replication on production cluster server; but i have little doubt what if replication will go down anytime..will it fail over cluster.

    I believe it should not failover, but are there any situation where it can?




    Deepak Kumar
  5. Argyle New Member

    A replication failure with a snapshot or distribution job will not cause a failover. Only a problem with one of the nodes like hardware or network will cause a failover of the sql service and it's disks. If the specific sql instance that is being failed over has replicaton configured on it, it will follow the instance over to the other node.
  6. deepakontheweb New Member

    Thanks Argyle;

    Can forum also help me for choosing hardware for cluster server with replication, which server model is better form UNISYS, DELL and IBM.

    what kinda cache i should use more for better transactional replication perfomance, if i have raid 10 and 1 for data and log files on 10K speed disks.



    Deepak Kumar

Share This Page