some queries regarding Recover Database using backup controlfile....

hi,

Assumming I have a db in archivelog mode with auto archiving...and i hv taken :

a)'alter database backup controlfile to ' controlfile binary backup AND

b)'alter database backup controlfile to trace' keep that recreate controlfile tracefile

...after that i have some transactions in the db, at this point if I've lost all controlfiles (including mulitplexing controlfiles) but all other oracle files are intact, how should i recover my database until point of failure?

1)should i restore my backup binary controlfile to the original controlfile location AND 'Startup Mount' then run a 'Recover automatic database until cancel' OR

2)'Startup Nomount' then run a trace file to recreate the controlfile
and then 'recover auto database until cancel' and then 'alter database open resetlogs' ??


Will I LOOSE any data that is still inside online redo logs??
any input will be appreciated.


thanks.

- Readup on backup and recovery concepts.
a) When you recover using "backup controlfile", you MUST open with resetlogs, there is no way around it. You can however feed the online redologs during recovery as if it were an archivelog, and thus minimize your data loss.
b) Alternatively, issue a create controlfile with NORESETLOGS option. Then recover all your archivelogs and issue a 'alter database open'. It should come up.

Either way will do the job.

Q for Axr: In your step 2 proc. above:

Scenario: control+db files gone..arch+online redo available!

Recovery proc:After restore of db files..
startup nomount
CREATE controlfile ... NORESETLOGS
RECOVER DATABASE
..Provide all the archived redologs it asks for
alter database open NORESETLOGS;

Tom(of Asktom): Confirms the above proc. saying:
you would recover the database and it would apply the logs.
the log sequences and such are "well known" -- they are just numbers..

My question: How/where are they "well known" as the controlfile has just been recreated although logically I get the fact that the redo chain here has not been broken and db need not be opened resetlogs? Is the redo chain recorded anywhere else..can only think of the view v$controlfile_record_section but does'nt it also read the current controlfile - or the alert log but what if you're trimmming it - or anything else I'm missing here?

Had posted this before - thread got closed so could'nt get any other opinions/info !!

:confused: :cool: :)

cruser -
Every datafile is associated with a SCN. SCNs are stored in file headers. When you mount a database, you are able to see the view v$archived_log. There you've got the first_change and end_change for every archive logfile. In addition, the fixed views v$recover_file and v$datafile track change#s as well.

Once you take a "consistent" backup of the database, and mount it and issue a "recover database;" - Oracle then looks at each datafile - figures out the oldest SCN. For eg, say data01.dbf was backed up before system01.dbf, and its SCNs are older than system01.dbf's - then the "recover" figures out the appropriate archivelog file to be applied based on the above view. Once you feed it, it merely applies the blocks over to data01.dbf. Then, it asks you for the next sequence# and so on. Once you go beyond the SCN's in system01.dbf, we're talking 'forward / future recovery' in some sense (for lack of better term). Then it is merely a question of asking for appropriate archivelog files based on the archive destination and file format specified.

That's what Tom probably refers to as "well known"..or so I guess. (I don't know what thread your are referring to..or the context)

Hope that helps.

But when u restore a "consistent" backup - all db files -> from say 3 days back, wont the info. in the views be from 3 days back and your scn's also from this point in time ?

Quote:

---

Originally posted by cruser3
But when u restore a "consistent" backup - all db files -> from say 3 days back, wont the info. in the views be from 3 days back and your scn's also from this point in time ?

---

Yes, they will be. Read my previous response again.

hi cruser,

my qn is if in my original scenario of loosing all current controlfile but i still have an old controlfile binary backup, in the event of recovery how does it know which is the latest archivelog/online redo log to apply? my impression is that the controlfile keep tracks of all the log sequences and therefore in the best knowledge to know which archive/online logs to apply? or am i missing something?

thks

Quote:

---

Originally posted by ngwh
my impression is that the controlfile keep tracks of all the log sequences and therefore in the best knowledge to know which archive/online logs to apply? or am i missing something?

thks

---

you are missing something. :rolleyes:

I tried..

Sure - but the backup is from say 3 days back - controlfile re-created say "today" - some log switching + archiving took place in the 3 days and are removed to tape and db files lost - how would the recreated cf know about this add. redo as the v$views, SCN's r from the 3 day old backup ?

For ngwh: What Axr has said for your problem will work !!
For Axr: Though I know the re-create cf will work if redo chain is not broken - I am not clear how the re-created cf will know of the redo chain from say a 3 day old backup to the point when failure occurred !!

:)