I've only had bad luck with the sqlnet.expire_time (unit is in minutes).
Seems dead detection is implemented on the server, not the client. If
there server determines a dead session it removes the shadow process
that supports it on the server. The client remains unchanged. If the
client hangsthe ping packets that are supposed to be send to the server
don't get there and your shadow process will be eliminated. The client
dead detection is based solely on the TCP/IP stack implemention. With
MS/TCP on NT, Clients are required to make specific calls via winsock
API to take advantage of socket timeouts. Oracle does nothing specific
to utilize these dead detection features on the client (they code to the
mose common API features available). So you if can't change the default
behaviour of socket connections you must live with what you have until
either MS or Oracle do something to support it. So Far I have tried
setting the keepAliveTime on NT, but was not successful in getting
results.
And so the saga continues....
Received on Fri Sep 05 1997 - 00:00:00 CDT
