Re: orasvr hangs on HP-UX 9.04

In article <2pnq3j$6kr_at_hermes.uni-konstanz.de>, timo_at_PROBLEM_WITH_INEWS_GATEWAY_FILE (Timo Salzsieder) writes:
> Michael Greenberg (michael_at_mcil.comm.mot.com) wrote:
.
.
> cause the error isn't reproducable. But we have an idea concerning the
> timeout mode of the orasrv-program. You can set the timeout time for the
> orasrv using tcpctl timeout <sec>. This is useful because some users ends
> their sessions without leaving e.g. sqlplus with the quit or exit command.
> The server listen to the connection although the connection doesn't exist in
> real. With the timeout set you can set the time the server wait until the
> session is killed not clearly. By default this time is 0 which means the
> server wait in an endless loop that explains the painful idle-time.

I think you are confused with timeout and keepalive. What you are describing is 'keepalive'. There were lots of discussion about keepalive a few weeks ago. Have a look at the previous postings on it.

Now, randy67_at_delphi.com is closer to the answer and wrote:

In article <pkwuUwT.randy67_at_delphi.com>, randy67_at_delphi.com writes:
> Eric Falkjar <falkjare_at_nbnet.nb.ca> writes:
>
> >So far I count 4 sites in diferent locations with the same problem.
> >Why has Oracle told me that the problem had never occured before?
> >I see that we are all using HP's. It doesn't have to do with the database.
> >Also too hard to duplicate so I guess Oracle wants us to go away.
>
> We have also had the problem with v1.x of orasrv on a v6.x oracle pc plat-
> form running sco unix, and on a v7.x oracle running on a sequent.
> sometimes on the sequent the orasrv sessions cannot be killed, and
> oracle has to be restarted. How does the timeout work? Does it mean
> that timeout=3 -- kill if a given user has not used it in 3 minutes?

From my recollection on SQL*Net's timeout. What timeout is used for is in establishing the connection from the client to the server. It will keep trying to establish the connection until it reaches the time specified by timeout.

You specify time out by 'tcpctl timeout n_sec' and check its value by 'tcpctl timeout'.

Note that there is another variable that should be set. This is the retry option. 'Retry' specifies how many times the connection should be retried if it fails. Now if RETRY COUNT = 0, then it will wait until timeout is received.

Now the ORIGINAL poster had problems with a HP-UX client to a HP-UX server, not a PC client to a HP-UX server. The problem is to do with the orasrv process chewing CPU cycles. I believe that setting either the 'retry count' or the 'timeout' on orasrv should solves this problem.

If I am wrong, please let me know and if it works for the 4 sites experiencing it, can you please post a summary to this group about the solution that worked? Most of the documentation can be found in the $ORACLE_HOME/tcp/doc directory.

ta
tony

-- 
 _____       ________ / ____ |Tony Jambu, Database Administrator
  /_  _        /_ __ /       |Colonial Mutual Invest Mgmt,Aust (ACN004021809)
 /(_)/ )(_/ \_/(///(/_)/_(   |EMAIL: TJambu_at_cmutual.com.au
 \_______/                   |PHONE: +61-3-2831639       FAX: +61-3-2831090