Re: Unable to shutdown the database
Date: Tue, 6 Nov 2012 14:29:52 -0800 (PST)
Message-ID: <f41ce9f6-2acc-40a7-b614-391a5a101508_at_i2g2000pbi.googlegroups.com>
On Nov 2, 8:04 pm, zigzag..._at_yahoo.com wrote:
> On Friday, November 2, 2012 3:46:22 AM UTC-4, zigz..._at_yahoo.com wrote:
> > I am on Oracle 11.1.0.7 on HP UNIX 11i. I have been using following procedure to shutdown database for a long
>
> > time:
>
> > 1. shutdown abort
>
> > 2. startup
>
> > 3. shutdown immediate
>
> > Today (after Sandy hurricane where my datacenter lost network connectivity), I noticed:
>
> > In the last step, database does not close, I have been waiting for 30 minutes and gave up. My database is only 50GB.
>
> > Any idea, why?
>
> > When I could not do step 3, I did again following steps:
>
> > 1. shutdown abort
>
> > 2. startup
>
> > Database comes up and I do not see any errors. I can access all my data. Can anyon eexplain, database will not shutdown without abort(why I have to wait forever, how I can troubleshoot).
>
> As I said earlier, because of Sandy storm, my site where database and app servers are installed lost wide area network connectivity. As a result I could not remotely logon to database server. Then inside the site, someone managed to connect using UNIX server console. Then the person was trying to shutdown oracle and restart it. We have scripts which were developed to be called from UNIx’s /sbin/rcx.d …when server reboots. Even though we were not rebooting the server, we were using those scripts to recycle oracle and shutdown was getting hung.
> Today, our WAN connectivity was restored, so I could logon to DB server remotely and wow I could do all the three steps in 2-3 minutes and shutdown immediate completed in a minute instead of “hanging” even after 30 minutes.
> May be, UNIX server was somehow trying to connect through wide area network while shutdown immediate was going on.. I do not know why it will need to go through WAN when Oracle is being shutdown…; but restore of WAN made all the difference. All my disks on db server are “local”; i.e., we do not use SAN in this site.
If you have an rman backup going from somewhere else, it may be possible for it to spawn parallel processes, which continue to try to communicate with the parent that is no longer reachable. These oracle processes would ignore a kill session command, requiring death from the OS. I know I've seen this locally, when the parent is accidentally killed, I'm speculating about WAN. But you could try it if you have a suitable test setup, and it could possibly apply to any parallelization operation.
> Thanks a lot for your insight on different ways of shutting down. I have read at few places, there is nothing wrong with shutdown abort (other than users uncommited work will be interrupted) and then doing startup and then shutdown immediate . You will not risk your database in anyway by doing 3 steps. Instead of just one step shutdown immediate.
Remember, most of those are simply quoting each other, some are "heman" implying you are a scaredy-cat to be afraid of shutdown abort and that it can't be a problem if they've never seen a problem, and some simply don't acknowledge bugs or extreme situations for educational purposes. The one you've pointed at is certainly out of date.
jg
-- _at_home.com is bogus. https://plus.google.com/photos/106910392693696947072/albums/5374880216169839969/5781744200831793442Received on Tue Nov 06 2012 - 23:29:52 CET