Thread: network failover configuration
06-21-06, 11:48 #1Registered User
- Join Date
- Feb 2004
Unanswered: network failover configuration
This was posted on Oracle Metalink, Lazy DBA, DBAClick, and DBForums. I appologize if you are reading this again.
I need a recommendation for network failover configuration. We're running Oracle 10gR2 (10.2.0.2) on Windows 2003 Advanced Server.
There are 2 groups of 5 database servers, one of which is intended to be the "production" group and the other a "backup" group. The 5 servers for each group are in the same rac, but the groups are located in geographically distant locations. These servers are used to query data which is loaded nightly on each group via SQL*Loader. The query application connects to the first server of each group ("server a") which contains frequently used tables and views using database links to the other 4 servers in the group ("server b", "server c", ..). The application is not aware of any servers other than "server a".
I would like to have the groups of servers failover as a group i.e. if "server c" in group 1 becomes unavailable then the application connects to "server a" in group 2 for handling queries and group 1 is considered offline. Failover of each individual server, i.e. "server b" in group 1 failing over to "server b" in group 2 has unacceptable response times for queries.
How can I configure failover so that connections are redirected to "server a" in group 2 if any server in group 1 becomes unavailable?It was working just 5 minutes ago - I promise !
06-22-06, 01:41 #2Registered User
The 5 servers for each group are in the same racThe query application connects to the first server of each group ("server a") which contains frequently used tables and views using database links to the other 4 servers in the group ("server b", "server c", ..)
- Join Date
- Aug 2004
I would be interested in knowing if you ever find a solution on another forum. Could you please post the link here if it happens ?
rbaraerORA-000TK : No bind variable detected... Shared Pool Alert code 5 - Nuclear query ready .
06-22-06, 08:35 #3Registered User
- Join Date
- May 2004
- Dominican Republic
Perhaps by RAC he means on the same exact location (for example, container), not part of a cluster.
What I see you can do is to make server a of each group a RAC database, that way, when one of either fails, it will failover to the other and thus changing groups (since you say server a on each group controls query/data flows). Of course, this will shorter your availability since you're going to only host two nodes, instead of 10. Also, RAC is intended for "in-house" nodes (same location) I dont know whether it will work ok in your configuration, since you say they're location geographically differently. You must have strong interconnects for that..