Re: new posts to an old issue

yls177 wrote:
> 
> http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&threadm=3f927ab3%240%243699%24afc38c87%40news.optusnet.com.au&rnum=8&prev=/groups%3Fq%3Dyls177%2Bgroup:comp.databases.oracle.server%26hl%3Den%26lr%3D%26ie%3DUTF-8%26group%3Dcomp.databases.oracle.server%26scoring%3Dd
> 
> hi, from the above... i cant post to that again. so have to link these
> two together...
> 
> basically... its like this
> 
> recently.. my colleagues have come up with 2 shutdown procedures and
> are as belows
> 
> 1)
> alter system switch logfile
> alter system checkpoint
> shutdown abort
> startup restrict
> shutdown immediate
> 
> 2)
> alter system switch logfile
> alter system checkpoint
> shutdown immediate
> 
> are there any major differences between them?
> 
> we noticed that (1) shutdowns much faster than (2).. this is because
> of the abort statement?

Think of it this way:

You're making a phone call.  Issuing the shutdown command basically
stops all new connections, so thinking only about the existing
connections:

Your version 1 .... someone comes along and takes a cleaver to the phone
cable (shutdown abort). Then gets out the repair tape (startup restrict)
and tells everyone to hang up.

Your version 2 .... someone comes along and says 'finish your sentance
and hang up now, please' and waits patiently until everyone has said
their good-byes and has hung up.  

If someone has walked away from their terminal (or hasn't stopped the
intelligent agent), #2 could take a LONG time.

Although this seems to verge on myth - in general, it's not considered
good practise to issue a shutdown abort on a regular basis.  However, if
you do issue a shutdown abort, it is good idea to do a startup restrict
and shutdown as soon as practical thereafter.  The abort leaves things
in a moderately confused state that requires instance recovery, which
happens automatically and generally cleanly on startup.  (However, there
is a small, very incredibly tiny, even miniscule, chance that the
database could become bunged if a disk bump happens before the startup
command.)

(Did I get that right?  I think it was HJR & Quarkman that helped me
understand this better <g>)
/Hans