I would wager that the problem is that database 1 & 2 are being resolved through different paths at the TNS level. Perhaps one is using a local tnsnames.ora file and another is going to the DNS server? If you're trying to resolve database 1 via DNS, discovering that it's not in the DNS server, then resolving it through another approach (Oracle Names, tnsnames.ora, etc), that could account for the difference.
Along the same lines, I'd check to see whether the host name for either database is in /etc/hosts (or the Windows lmhosts file). Defining the values there bypasses the call to the DNS server.
At 12:58 AM 2/3/2003, VIVEK_SHARMA wrote:
SITUATION - With DB Server CPU Idle being 0 % & load average 60-70 % :-
tnsping Connect to one of the Databases (1) takes about 10,000 ms
truss -fdD tnsping <connect string>
29504: 20.0006 18.8113 read(3, "\0 K\0\004\0\0\0 "\0\0 ?".., 2064) = 75
Thus 18.... seconds taken as shown above
Whereas to another Database (2) tnsping takes only 10 ms on the Same machine
DETAILS :-
0) 1 GBPS network path exits from APP Servers to the Database Server.
1) Database 1 Has about 3500 Oracle processes connecting to it
Database 2 Has about 1500 Oracle processes connecting to it
2) Both Databases have Different ORACLE_HOMEs
3) sqlnet.ora of Database 1 :-
AUTOMATIC_IPC=OFF
TRACE_LEVEL_CLIENT=OFF
TRACE_LEVEL_SERVER=OFF
TNSPING.TRACE_LEVEL=OFF
NO sqlnet.ora exists on Database 2
4) NO process.dat , regid.dat exists in ORACLE_HOME/network/log Dir of ORACLE_HOME for Both Database
5) Multiple listener processes are Running for BOTH Databases
Qs How may the issue be approached ?
Justin Cave
Distributed Database Consulting
