RE: 10g - Sqlnet and redundant oid servers
- From: Yong Huang <yong321@xxxxxxxxx>
- To: cjnewman@xxxxxxxxxxxxx
- Date: Wed, 5 May 2010 09:16:46 -0700 (PDT)
> We recently had an issue where clients were taking 8 minutes
> to login to the database; the primary OID server (server1 below)
> was down at the time, the secondary was up.
Chris,
I tried to reproduce this but can't. My test with 10.2.0.1 Oracle on Windows XP
shows that the name (connection string) is still instantly resolved and tnsping
returns OK if the first OID server is unreachable (can ping, but OID port is
closed because there's no OID software installed) or if its name can't be
resolved by DNS. I also tried this with 11gR2 and 10.2.0.4 Oracle on 64-bit
Linux.
There's only one line in sqlnet.ora:
NAMES.DIRECTORY_PATH= (LDAP,TNSNAMES,EZCONNECT)
and three lines in ldap.ora:
DIRECTORY_SERVERS= (server1:389:636,server2:389:636)
DEFAULT_ADMIN_CONTEXT = "dc=<domain>,dc=..."
DIRECTORY_SERVER_TYPE = OID
Maybe your issue was about reaching a non-existing Oracle database server
*after* the name and detailed connection info was already fetched from OID? If
you set your DB server to HOST=123.45.67.89, TCP timeout takes effect (21
seconds on Windows and at least 3 minutes on Linux). See
http://yong321.freeshell.org/oranotes/TnsTimeout.txt
Yong Huang
--
//www.freelists.org/webpage/oracle-l
Other related posts:
