Post reply

Cluster replication across a firewall

  • October 20, 2005 at 4:34 am

    #91838

    Hi

    I have a database on a cluster server that needs to be replicated to a remote SQL server (not clustered) on the other side of a firewall.  I have had a rule set up on the firewall so that the SQL Server IP address on the cluster can communicate on port 1433 with the remote server, and vice versa.  I can set up the remote server as a subscriber, but when I run the snapshot, the Distribution Agent fails with the message "The process could not connect to Subscriber 'REMOTESERVER'."

    What I suspect is going on is that it is the node (or cluster) IP address that is trying to communicate with the remote server, rather than the SQL IP address.  Before I go to the trouble of asking for any other IP addresses to be allowed to communicate through the firewall, can anyone confirm my suspicion?

    Thanks

    John

  • October 20, 2005 at 6:04 am

    #598653

    Can the publisher ping the subcriber by netbios name? Go to your subscriber and issue the following.

    xp_cmdshell 'echo %computername%'

    select @@servername

    These should be identical.

    Do the same on the publisher. Again they should be identical.

    Then build an alias to the subscriber using client network utility so that the alias matches what xp_cmdshell 'echo %computername%' gave on the subscriber. The server name will be the fqdn to the subscriber or its ip address.

    --

    Hilary Cotter

    Looking for a SQL Server replication book?

    http://www.nwsu.com/0974973602.html

    Looking for a FAQ on Indexing Services/SQL FTS

    http://www.indexserverfaq.com

     

  • October 21, 2005 at 3:29 am

    #599042

    Hilary

    Thanks. On the subscriber, the %computername% and @@servername information were identical. On the publisher they were identical except that @@servername had \instancename on the end.

    I created the aliases as you suggested.

    However, it is still not working. I ran a Profiler trace while the connection to the subscriber was being attempted, and then did a search through the trace for the name of the subscriber. In nearly all rows where it was found, the value for hostname was the name of the cluster node and not the name of the virtual SQL server.

    This leads me to believe that if I get the IP address of the cluster node allowed through the firewall (remember at the moment that only the SQL IP address is allowed) then it might work. Would you agree with that?

    Thanks

    John

  • October 21, 2005 at 9:15 am

    #599161

    Just to add my 2 cents:

    I've always noticed (since working in a clustered replicated environment) that the replication agents show up as the node name instead of the sql host name.  So I suspect that your reasoning is correct and your plan should work.

     

    Dylan Peters
    SQL Server DBA

  • October 28, 2005 at 2:08 am

    #600822

    Thanks, folks.  I had the extra rules put on the firewall and now it works as it should.

    John

Viewing 5 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic. Login to reply