Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Standby: Managed or Automatic?

Re: Standby: Managed or Automatic?

From: Joel Garry <joel-garry_at_home.com>
Date: 20 Feb 2004 14:28:43 -0800
Message-ID: <91884734.0402201428.1d5841af@posting.google.com>


joel-garry_at_home.com (Joel Garry) wrote in message news:<91884734.0402171725.360cb83a_at_posting.google.com>...
> bandar_at_gmx.net (bandar) wrote in message news:<e623b4e7.0402161337.5a1225cd_at_posting.google.com>...
> > ITAPORT06 <mokat67_at_hotmail.com> wrote in message news:<a8qv209fcqvbpnccq6b3or6msspul7oklk_at_4ax.com>...
> > > I work with Oracle 8i 8.1.7.4 on WINNT & 2K
> > >
> > > I did setup a Standby database envirnoment and found out (that for
> > > several reason which I still have to reasearch) that not all archive
> > > loggings where send to the the Standby server.
> > >
> > > The Standby server didnot apply the missing loggings which I shipped
> > > over manually. Server was in Managred mode.
> > >
> > >
> > > I shut down the server and brought the server back in Automatic mode:
> > >
> > > RECOVER AUTOMATIC STANDBY DATABASE
> > >
> > > All manual shipped loggings were applied succesfully.
> > >
> > >
> > > Mine question is:
> > >
> > > Is it not handier to make use of the Automatic option instead of the
> > > managed?
> > >
> > > The loggings wil be shipped by the NET8 configuration and if
> > > something will go wrong then there will be a synchronisation job
> > > running which willl ship the logs that failed to be shipped earlier
> > > by NET8 (This can happen because of a network errors etc etc)
> > >
> > > A scheduled job on the standby server will apply the logs by
> > > activating the Recover Automatic standby option every X minutes
> > >
> > > The reason why I want to use this method is:
> > >
> > > We do not have the resources to check the oracle servers frequently
> > > if all loggings have been shipped to the standby server. If one of the
> > > logs does not have been shipped then we will find this out much later.
> > > I try to prevent this by adding also the synchonisation job and
> > > making use of the Automatic option.
>
> You have to check _more_ if you are using your own scripts. You are
> basically reinventing the wheel. Granted, the wheel has bugs, but you
> have to deal with even more with your reinvention. What do you do
> when Oracle starts multiple archivers writing multiple files?

I'm going to disavow my previous post. Since there is no way to prevent the managed recovery standby from affecting the production instance in 8i, bottom-line, that is unacceptable. With your own scripts, you can completely disconnect the standby from the production.

As it is, the archiver writes 1M chunks to each of the destinations in a round-robin fashion, and waits for all to be completed before saying the archiving is finished. There are several circumstances involving the network or standby hardware that can cause the production archiver to think it is never done. This can stall the production db, or in extreme cases, require a resetlogs.

I have personally had two of these circumstances in two months, which is two too many.

9i is somewhat better, but still not there. There are additional parameters to deal with these issues, but they don't necessarily work in earlier versions. Support tells me there is an underscore parameter in 9.2.0.5 (!) and feature in 10g that allows keeping archives locally for subsequent shipment to the standby.

>
> If you are bothering to write a script, why not write one to check
> that things are working and notify if not?
>
> > >
> > > My questions:
> > >
> > > 1) Is this method ok or do you think it is still beter to go for the "
> > > managed " method. Take as fact that we can not check the server
> > > frequently for missing shipped logs and that we have network problems
> > > now and then.
>
> Yes, I grant the network problems, but I have real problems with not
> checking.
>
> I agree that the 8i method of dealing with network problems is
> inherently flawed. One way to deal is to allow lots of log (and
> perhaps local arc destinations) to give time to deal with the
> problems.
>
> > >
> > >
> > > 2) What are the real pro's for making use of the Managed method.
> > >
> > > Thanks
> >
> >
> > Standby using Oracle 8 is pretty much working with scripts and you
> > have to check your logs frequently independently of how you configure
> > it.
>
> ??? He's using 8i.
>
> >
> > I would try an upgrade to Oracle 9i with the Dataguard solution, that
> > could handle it a little better but you would still need to check your
> > logs frequently. Dataguard doesn't work with the Standard Edition, you
> > would need to upgrade to Enterprise ($$) if you are not already on it.
>
> Agreed.
>
> >
> > Alternatively think about third party products. I know that a prodcut
> > called "DBShadow" is handling log shipping very well.
> >
> > Good Luck.
>
> jg

jg

--
@home.com is bogus.
Yep, every year.  http://www.detnews.com/2001/livingston/0101/07/e05-172555.htm
Received on Fri Feb 20 2004 - 16:28:43 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US