Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Help required on Data Guard
"Howard J. Rogers" <hjr_at_dizwell.com> wrote in message news:<41abf6d3$0$12876$afc38c87_at_news.optusnet.com.au>...
> Mark Bole wrote:
> > Howard J. Rogers wrote:
> >
> >> Joel Garry wrote:
> >> [snip]
> >>
> >>>
> >>> RAC is, as you say, for a single database. Data Guard uses log files
> >>> to replicate the transactions to another database.
> >>
> >>
> >>
> >> Minor quibble: it actually uses processes, either LGWR or ARCn, to
> >> trasport redo. No files actually get shipped between machines, as such.
> >>
> >> HJR
> >
> >
> > When quibbling, are you talking about RAC or Data Guard?
>
> Data Guard: the one where what it uses is actually mentioned.
>
> > More specifically, in the case of Data Guard, are you saying that no
> > matter how far beind, the primary database, once it notices it can talk
> > to the standby again after an interruption, actually re-reads the
> > previously locally archived redo logs, and then sends the data, as
> > re-read, to the standby via log transport, where the redo log file is
> > then re-instantiated on the fly?
>
> Well, ordinarily, if archiving to a destination is deferred, and then
> re-enabled, no catchup of any kind is performed, and it's up to you to
> plug the gaps, and that would involve manually copying log files.
>
> But with Data Guard, you can configure a 'FAL Server' and a 'FAL
> Client', and additional processes are then responsible for doing exactly
> what you've described: mining files on the FAL Server, and shipping the
> extracted redo to the FAL client. I can't recall, however, whether that
> capability first made an appearance in 9i Release 2 (I believe so), or
> in 10g.
Well of course I was oversimplifying for the OP, but I didn't say the files were copied over. Your quibble is correct. But they do wind up there, right? I have to be careful about talking about FAL, because I'll start ranting about how it doesn't work the way one would hope, ie, up until latter-day 9i's, a network error can kick back through the FAL process and stuff your production instance bigtime.
>
> That, at least, is my understanding -though I note the 10g doco talks
> about "by automatically retrieving missing archived redo log files from
> the primary database". So even it talks about copying *files* across...
> but then quite a lot of the doco uses language loosely.
>
> When you have diagrams showing Oracle background processes talking to
> each other, it's a fair bet that they aren't having a chat about doing a
> simple 'cp'!
Frankly, for 8i, they should have. That's what many people's scripts do. But I'm sure everything is fine now.
>
> > Hmm, I'd never thought of that. But of course, that *would* be much
> > more portable across OS platforms.
>
> Precisely.
>
> Regards
> HJR
>
> > I did notice that the timestamps of
> > 'catch up' log files were contemporaneous with the catch-up, but I
> > thought that was just a sloppy way of copying files (for example, not
> > using the equivalent of "-p" = preserve option).
> >
> > -Mark Bole
> >
> >
jg
-- @home.com is bogus. Email is for the elderly: http://english.chosun.com/w21data/html/news/200411/200411280034.htmlReceived on Tue Nov 30 2004 - 16:36:06 CST