HP-UX

I have a .net application that works with an Oracle database in HP-Unix.

Problem: .net application creates Oracle processes that don't close. By default Oracle can work with 200 active processes and after that crashes. I doubled this number and I don't have to many problems.

I want: to close this processes.

Obs: the programmer tell me that he closes the .net connection after he sends an update/delete/insert to the Oracle however not all of this processes close by themself.

Any idea?

Thank you!

Im old and half stupid... so forgive my silly question:

What is running where? I dont know what .net application is and havent seen such thing installed on HPUX ( Its Microsoft stuff no for MS OS no?).
I understand you have a win server with a .net application accessing a HPUX server being a rdbms server is that right?

You are correct, I have a .net application on a Windows machine that is accessing the Oracle under HP-Unix.

By the way, sometimes HP-Unix closes the processes, sometimes it doesn't (take a look below - last line):

root:/ ps -eaf|grep oracleTRLV |wc
36 324 3178
root:/ ps -eaf|grep oracleTRLV |wc
36 324 3178
root:/ ps -eaf|grep oracleTRLV |wc
36 324 3178
root:/ ps -eaf|grep oracleTRLV |wc
37 333 3248
root:/ ps -eaf|grep oracleTRLV |wc
38 342 3357
root:/ ps -eaf|grep oracleTRLV |wc
38 342 3357
root:/ ps -eaf|grep oracleTRLV |wc
38 342 3357
root:/ ps -eaf|grep oracleTRLV |wc
39 351 3426
root:/ ps -eaf|grep oracleTRLV |wc
24 216 2220
root:/

Now it makes a little more sense to me...
The issues I see can happen on your HP will be
Users ( oracle ) cant connect anymore, and in worst case you loose also the connection...
Only a reboot can solve at this point (very rare...).

I will go and look if I can find you a ( live) sample...

The problem is that Oracle can work with X (by default X = 200) number of processes and after that it crashes.

The question that I have is this: how is it possible for HP-Unix sometimes to close the processes (see below) and sometimes not to?

root:/ ps -eaf | grep oracleTRLV | wc
38 342 3239
root:/ ps -eaf | grep oracleTRLV | wc
19 171 1909

No...
But here is things I would look at:
My experience with MS servers accessing UNIX boxes is MS parameter equivalent to Novell's Watchdog is a potential culprit:
The windows server has a connection timeout far too low and so believes it has lost a connection, closes certainly on his side and opens a new one on the poor UNIX which now has 2 opened connection with only one active the timeout on unix side is quite some time and if the windows server goes bezerk, then it will exaust all your connection resource (sockets).
You will therefore have to check with netstat command:

Code:

oke2:/home/vbe $ netstat -a|grep WAIT|wc -l                                           
119

The ones that are your concern are high FIN_WAIT_2 and maybe some TIME_WAIT

my 2 cents...

All the best

What OS version and plateform?