Re: New Solaris announcement

From: Noons <wizofoz2k_at_yahoo.com.au>
Date: Sun, 14 Oct 2012 14:28:09 +1100
Message-ID: <k5dbga$ncr$1_at_dont-email.me>



John Hurley wrote,on my timestamp of 14/10/2012 8:49 AM:

> # What, no kill -9?
>
> You probably know this already but with recent releases of Oracle the
> "kill session" or "disconnect session" usually work well and quickly.
>
> Only in rare cases of processes hung and waiting deep in oracle
> ( usually/always calling into os services ? ) is the kill -9 required.

Theoone I've found invariably makes me reach for the kill -9 gun is sessions using db-links to weird and wonderful places such as Postges, DB2/AS400 and such. Invairably if the other end goes mute, the Oracle side won't budge until I shoot the background process and the db has a chance to recover the session.

> Unfortunately when the kill -9 is required the system is usually in
> such big trouble that unless you have some automated scripts doing the
> scalping ( and all the inherent dangers of using such a draconian
> solution ) ... it may be bye bye instance.

I must admit since 10.2.0.3 I have yet to experience a hung instance anywhere. At least from unknown reasons.
Of course: when a "SAN expert" decides to remove all my LUNs from the SAN cache "because the db is killing all other I/O" and other such pearls of idiocy, things can go bump. But I don't call that db hangs. Although the word "hang" then gets mentioned in relation to said "experts"... Received on Sun Oct 14 2012 - 05:28:09 CEST

Original text of this message