Re: INIT forks too many ORACLE instances...

In article <61370_at_rnd.STERN.NYU.EDU>, rmadhava_at_rnd.STERN.NYU.EDU (R. K. Madhavan) writes:
|> Hi Folks...
|>
|> I am posting this for my brother who does not have access to the net.
|> He is working abroad and so if you need further details, I have to get
|> on the horn over this weekend. Please email me any suggestions as I
|> am not a regular reader of the newsgroup (Shame on me! But as my
|> doctoral-dissertation adviser puts it copious free time awaits me as
|> soon as I make tenure!)
|>
|> Here goes...
|> ----------------------------------------------------------------------
|> We are facing a very peculiar problem with our Project
|> Implementation... Of course , I am not directly involved or assigned
|> to the problem but I am was very much interested and am following it
|> closely...
|>
|> When the Unix system is booted the "Init Process" brings up the Oracle
|> Database i.e 4 oracle processes are fired The parent id of these are
|> 1. After that the system is ready for use.. i.e the users logs in and
|> starts using the application.(Oracle Financials).. Now , irrespectve
|> of the load on the system , this peculiar thing happens.. A Unix
|> Process with the owner as the application login, "orapps" in this case
|> and the parent id as "1" gets created at regular intervels... and
|> these processes never die , but chew up a whole lot of CPU time.. The
|> process name (COMMAND) is something like "oracleAU01 P:4036,5,8" i.e
|> it is a call to ORACLE... (AU01 is the name of the database instance)
|> And the System gets very very slow as the no. of these processes
|> increases... As an interim solution they either kill these or do a
|> shutdown and bring up the system again..
|>
|> The questions are:
|> =================
|> Why are these processes forked by init (PID 1)?...
|>
There is a call to start Oracle in /etc/inittab.

|> Will "init" fork even after the system has been booted and is
|> running..?

Yes, if the line to start Oracle includes the flag "respawn".

|>
|> Most importantly what can I suggest (and earn a promotion!!) to arrest
|> this phenomenon?

Don't start Oracle from inittab. Try starting from rc.local (on BSD based unix boxes) or as part of the rc2 scripts (on system V - for example /etc/rc.d/8/userdef on SCO).

|>
|> I looked thru the log files to which Oracle Spools all its messages.
|> There were several error messages saying "Semop Error, unable to
|> increment semaphore". This error could possibily be because too many
|> processes (because of the above problem) were fired and oracle ran out
|> of semaphores...
|> ------------------------------------------------------------------------
|> Any takers? I hope so...
|>
|>
Hmm. Probably caused by all the startups. If it doesn't go away, call Oracle support and winge.

Hope this helps,

Regards,

Simon.
(Just my opinion, and not anyone elses, probably...) Received on Mon Apr 11 1994 - 14:25:21 CEST