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: Which Archive-Logs are to delete?

Re: Which Archive-Logs are to delete?

From: Frank an der Heiden <fh_at_energotec.de>
Date: Sat, 7 Oct 2000 17:58:05 +0200
Message-ID: <8rnh5i$738$1@oxygen.technet.net>

hi,

"Howard J. Rogers" <howardjr_at_www.com> schrieb im Newsbeitrag news:39deea2e_at_news.iprimus.com.au...
> The short answer is that you need to retain all the archives from the time
> that the last complete backup cycle was started. To this end, it makes no
> difference whether a 'backup cycle' is accomplished by copying all
 datafiles
> in a single 3-hour maintenance window, or if it takes you six weeks to do
 an
> entire backup, one tablespace at a time. It's the time at which you START
> the backup cycle that is critical.
>

Normally I do every day a complete online backup of ALL tablespaces of the database.
(the whole database is only about 800 MB) After the backup I checked the column CHECKPOINT_CHANGED# in V$DATAFILE! The requested archivelog defineately had an LOWER sequence number (checked in V$ARCHIVED_LOG) than the sequence number related to the checkpoint written for all datafiles in V$DATAFILE.

This is what I can't understand. Why did oracle request this file? I'am running Oracle 8.1.5 Server for NT 4.0, SP6a

It also happend not during a "normal" backup-recover, but after rebooting the NT server workstation after updating a NIC driver. The database did not open with the message that a media recovery for a datafile is necessarry.

> However, this has come up before: whilst it is technically possible to
> delete prior archives, you would be mad to do so (because unless your
 middle
> name is 'The Lord Almighty", you can't guarantee that the new cycle backup
> is going to be fine -so when I get the failure on the 2nd Tuesday, it's
> quite possible to find that the prior cycle is dodgy in some way, and you
> are forced to go one cycle further back than expected -which, of course,
 if
> you've been too eager isn't there any more.
>

You are completely right and I already planned to it in this way.

> Incidentally, it appears from the description of how you perform your
> backups that you FIRST delete the prior archives, and THEN do the new
> backup. This is extremely bad practice and renders you in a window of
> vulnerability. You should only ever delete archives when your new
 complete
> backup has successfully been taken (and has been verified as OK).
>

Maybe, my description was not so good, sorry. For sure I do not physically remove these archivelogs before the backup did perform sucessfully. I only rename these files in order to be able to identify
OLD archivelogs (which are not necessarry any longer) and NEW written ones.

> Regards
> HJR
> --

Thanx for prompt reply
Frank

> --------------------------------------------------------------------------
> Opinions expressed are my own, and not those of Oracle Corporation
> Oracle DBA Resources: http://www.geocities.com/howardjr2000
> --------------------------------------------------------------------------
>
> "Frank an der Heiden" <fh_at_energotec.de> wrote in message
> news:8rktii$5eu$1_at_oxygen.technet.net...
> > Hi folks,
> >
> > I have a problem with an Oracle database (Version 8.1.5 NT)
> > running in archive_log mode.
> >
> > In order not to waste harddisk-memory, I want to delete archivelogs
 which
> > are not needed for the recovery procedure.
> >
> > The normal way was, I deleted my archive-logs, started the backup with
> >
> > alter system switch logfile (which does an 'alter system checkpoint')
> > alter tablespace xxx begin backup
> > copied the related datafile to my backup-destination
> > finished the backup with alter tablespace xxx end backup.
> >
> > Now it happend two times during some tests, that the recovery of my
 backup
> > failed, because oracle did await an archivelog which had an 2 days
 EARLIER
> > date
> > then the date when the backup was performed.
> > For example the sequence_number of the first archivelog AFTER the backup
 was
> > 120,
> > but Oracle requested an archivelog with the sequence number 117 (was
 written
> > 2 days before
> > but deleted when the backup started => memory reasons).
> > During the backup ALL datafiles where backed up!
> >
> > How can this happen?? Why does oracle request this old archivelog? After
> > performing
> > an 'alter system switch logfile' all transactions should be in the
> > datafiles, or not??
> >
> > OK, I could even store also the old archivelogs, but I do not understand
 why
> > this
> > happend.
> >
> > No I try to find out, which archive-logs are really necessarry for a
> > successful recovery
> > of my backup-set by using this join:
> >
> > SELECT A.SEQUENCE#
> > FROM V$LOG_HISTORY A, V$DATABASE B
> > WHERE A.FIRST_CHANGE# >= B.ARCHIVE_CHANGE#;
> >
> > Is this the correct way, has anybody experience or tips for an easier
 way
 to
> > find
> > out, which archive-logs are possible to delete?
> >
> > Every help is very welcome because this really is a serious problem for
 me
> >
> > Thanx in advance
> > Frank
> >
> >
>
>
Received on Sat Oct 07 2000 - 10:58:05 CDT

Original text of this message

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