September 26, 2005 at 2:37 pm
Hello,
Every now and then when we have problems with our MS cluster, we find that linked servers that we have defined to external Oracle databases start failing. Linked servers to MSSQL (v6.5+) continue to operate. 'Moving' the database to another node of the cluster consistently corrects the problem.
One note: We have two instances running on the cluster. When both the default instance and a named one are running on the same node, we only have connectivity issues with the default instance. Linked servers to Oracle from the named instance continue to work while identical ones in the default instance fail.
Has anyone seen behavior similar to this before? Is there an OS process that I can restart that might restore my linked server connectivity?
September 27, 2005 at 12:35 am
Are you using Oracle RAC or is there more than one (Oracle) listener ?
What version is the Oracle Client ?
What problems are you having with the MS cluster ?
DB
The systems fine with no users loggged in. Can we keep it that way ?br>
September 27, 2005 at 10:57 am
Sounds familiar. We had Oracle 9i driver on a cluster and had to move it due to heavy usage of the linked server. Even with a -g startup paramenter (to give the provider more memory) didn't help. Failover freed up the memory temporarily until it ran out again--vicious circle.
I ended up moving the linked server to another box using the Oracle 8 driver, and using OpenQuery statements (which ran faster). I set the startup option at -g690, but the amount depends on how much total RAM you have.
Try using the -g switch in the SQL Server startup first. See BOL "Using Startup Options". Good luck.
September 27, 2005 at 12:32 pm
The event log on one of the nodes simply states 'Cluster resource 'SQL Server' failed. The database instance restarted at that time and we have not been able to connect to Oracle via a Linked server since. We are not using RAC and the Oracle client is 9.2.0.1. I'll look at using the -g switch.
Thanks!
September 27, 2005 at 12:40 pm
If that doesn't help, think about downgrading the Oracle driver to version 8. The 9i driver still caused issues on our second server too.... the Cluster nodes are on Windows 2003 and so is the other separate server.
September 28, 2005 at 5:51 am
Have you verified that the Oracle client is installed on all nodes of the cluster? Is the client the same version? Are the TNSNames.ora and SQLNet.ora files the same on all nodes?
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply