Re: alter system checkpoint in hot backup script. Why

"Burt Peltier" <burttemp1ReMoVeThIs_at_bellsouth.net> wrote in message news:suzvb.21735$ow5.21074_at_bignews2.bellsouth.net...
> Thanks for the great reply. Also, sorry this is getting to be a lot longer
> post than I wanted, but it is necessary I think.
>
> I think I understand the issue that apparently I was not aware of before
> (needing a checkpoint in our particular case).
>
> Please correct me if I incorrectly summarize for me or anyone else looking
> at this (also have 1 more question at the end of my post).
>
> Note: I don't think I am being paranoid.

I used that word in fun (which is probably bad taste...). I didn't mean it as a criticism. Production data deserves a bit of paranoia anyway!

>I am trying to cover the particular
> case of loosing EVERYTHING including REDO logs and subsequent archived
redo
> logs after the last night's hot/online backup.
>
> Summarizing: We need to do the "alter system checkpoint" after a
hot/online
> backup if I want to ensure uncommitted REDO entries are written from
buffers
> to the REDO log disk files. This would then allow the "alter system
archive
> log current" to copy off the REDO log to the archived redo log dest. This
> archived redo log file will NEED to be included in our hot/online backups,

Spot on so far.

> IF we want to recover the database to a consistent state from the last
> night's backup and only loose 1 day's worth of data (data since the last
> backup). Also, if we had the previous night's backup and archived redo log
> during THAT backup, then we could restore from that backup for a
consistent
> state. But, then of course, we would loose 2 day's worth of data.

I'm reading this at a fair lick, but I can't see anything wrong with the logic so far.

> Note: I don't want to go off on another issue, but this particular
database
> is completely stored on 1 RAID partition (not my choice and out of my
> control). So, if we loose the RAID partition, we loose ALL Oracle files
and
> the Oracle software too. Also, no one is questioning performance
> implications (because of 1 RAID partition) or the loss of 1 day's data.
>
> Also, there is one slight complication for me. The hot/online backups are
> using RMAN full backups on weekends and then incremental nightly backups.
> Note that this is an Oracle8i 8.1.7.4 database on W2K.
>
> I am not as familiar with RMAN backups (especially incremental). Anyway,
it
> does NOT seem that this should make any difference (still need the
> checkpoint and archive log current commands).

Off the top of my head, I can't see that using RMAN makes any difference. Only one thing: whilst the 'alter system checkpoint' is still required ("required" in the sense we've already discussed), the archive log current isn't, because RMAN does that for you for free in 9i. But as you are using 8i, then yes, even that bit remains the same... you'd still need to see it at the end of your script.

Regards
HJR Received on Sat Nov 22 2003 - 00:23:11 CST