Re: Oracle 8i (8.1.7) problem with a lot of simultaneous connections -> ORA-12541 No listener available

As you know the other server listener is up, it means something is busy. THe client timedout waiting for a response. Whether the client missed a message, the network bottlenecked or the remote server bottlenecked (not giving the listener enough time) is not known. I have also seen the message and your symptoms when there are other network problems (100MB connections thru 10MB ports, routing tables scrammed and rebuilt, etc.)

When the threads fail, is it always the last of the test pack or is it random within the group?

"Tomislav" <shaka_at_envox.hr> wrote in message news:<MRLN4495E1EDF_at_merlin.envox.hr>...
> I have a multithreaded application written in Windows 2k/C++ that uses OCI
> to
> access a remote Oracle 8i database. Both versions of Oracle are 8.1.7, on
> Client and Server side.
>
> I use a single OCIEnviroment handle for all threads and I initialize &
> destroy it only once with OCI_THREADED flag.
>
> Each of my threads opens a connection and executes an Select statement over
> and over again (infinitely - in this test case).
>
> When I start 16 threads at the same time, they wait a bit for opening the
> connection to the remote Oracle server (in the 100 Mbit LAN), after that
> they start
> executing simultaneously without any problems Select statements and fetching
> small resultsets.
>
> Different case is when I start like 32 threads. Then at database opening
> time some
> of them (like 6) produce "ORA-12541: no listener available" error message
> when trying to open database connection. When I try to open the connection
> again in those threads, it succeeds.
>
> When I start 64 threads, 27 of them produce "ORA-12541" and 15 of them
> produce "ORA-12541" even the second time they try to open connection as
> well. Third time they succeed and after that all of them (all 64 work at the
> same time).
>
> Is this a message of type "Network busy" or "Client busy" or whatever it
> means?
>
> Why isn't this explained better in documentation or signaled better through
> OCI return codes?
Received on Mon Jan 14 2002 - 14:32:43 CST