Checkpoint Recovery

Checkpoint Recovery

Post by Sympatico News Serve » Sat, 20 Feb 1999 04:00:00



Hi,
    We are trying to simulate a disaster situation where we recover from a
full system backup say for a friday night and then recover from checkpoint
tapes up to the day of the disaster, say the Wednesday of the next week.  We
are having problems rollforwarding the checkpoint from what seems is that
the checkpoint sequence number is higher than what is in the .cnf database
file from the full system backup.  We are using the rollforward +c option
and restoring from mag tape.

    This seems a little crazy not being able to rollforward from a
checkpoint where the checkpoint number is greater than what is in the .cnf
file.  Is there anything we can do?

Kyle Hanson

 
 
 

Checkpoint Recovery

Post by Michael L » Sat, 20 Feb 1999 04:00:00



Quote:>Hi,
>    We are trying to simulate a disaster situation where we recover from a
>full system backup say for a friday night and then recover from checkpoint
>tapes up to the day of the disaster, say the Wednesday of the next week.  We
>are having problems rollforwarding the checkpoint from what seems is that
>the checkpoint sequence number is higher than what is in the .cnf database
>file from the full system backup.  We are using the rollforward +c option
>and restoring from mag tape.

>    This seems a little crazy not being able to rollforward from a
>checkpoint where the checkpoint number is greater than what is in the .cnf
>file.  Is there anything we can do?

>Kyle Hanson

Kyle,

The proper backup procedure for an Ingres database is not explicitly explained
in the manuals.  You can gleen it by reading every word of every manual ;-}

The proper backup procedure is:

        1. run "ckpdb yourdb" and put the output to safe media
        2. backup the dmp and jnl directories for yourdb to safe media
        3. backup the output of "infodb yourdb" to safe media

Without all this information, you cannot recover from a disaster (where more than
one device fails).  Simply put, rollforwarddb REQUIRES the checkpoint information
to recover a checkpoint.  This information is written to the aaaaaaaa.cnf file
in both the default data locate and the dmp file directory, but it is written
AFTER ckpdb completes its backup operations.  So the information on your checkpoint
tape is not sufficient.

Even with this information, you need to regularly backup your journal files
should you wish to recover to a specific period in time.

Lastly, if you haven't safely configured your installation accoring to CA
guidelines, you cannot recover "up to the minute" if a single device fails.

Years ago, I presented papers at CA World about backup and recovery.  Would
anyone be interested in attending such a session this year?  I could redo
my paper and add some new information for OpenIngres, etc. and present it
again this year.


Caribou Lake Software  http://www.cariboulake.com  Java, Oracle, Ingres
Minneapolis, Minnesota (612) 323-9713              

Java: It's better than cool.

 
 
 

Checkpoint Recovery

Post by Kelly Youn » Sat, 20 Feb 1999 04:00:00




> >Hi,
> >    We are trying to simulate a disaster situation where we recover from a
> >full system backup say for a friday night and then recover from checkpoint
> >tapes up to the day of the disaster, say the Wednesday of the next week.  We
> >are having problems rollforwarding the checkpoint from what seems is that
> >the checkpoint sequence number is higher than what is in the .cnf database
> >file from the full system backup.  We are using the rollforward +c option
> >and restoring from mag tape.

> >    This seems a little crazy not being able to rollforward from a
> >checkpoint where the checkpoint number is greater than what is in the .cnf
> >file.  Is there anything we can do?

> >Kyle Hanson

Kyle,
   On the rollforwarddb command, try using the parameter "#cnnn"
where nnn is the checkpoint number that you are trying to restore.
For example: rollforwarddb +c +j "#c123" dbname
--
Kelly Young
Maricopa Community Colleges

 
 
 

1. Checkpoint Recovery

Does Infomix provide me with the ability to specify a particular transaction
or checkpoint to be used in rollback or recovery processing.  My questions stem
s from a user scenario, where the user has requested a series of database
processes accidentally.  I would like to be able to actually re-create the
database at the point before the user actions.

Associated question would be how to I tie the user request to a database
state?  From a client workstation?

Thanks in advance...
 ___________
|           |   Regards,
|            :      Steve Malme
|   Fargo ->*|      Great Plains Software

2. SQL Agent - Job Cache error

3. query large text fields more than 256 characters in stored procedure

4. Strange influence of recovery interval on checkpoint (SQL SERVER 2000 SP2)

5. Help needed: Why this java code does not work?

6. Btrieve 6.15 errors

7. checkpoint and recovery interval!

8. Recovery considerations of checkpoint vs. log switch

9. Contents of database dump files / effects of checkpoint on recovery

10. Checkpoints/disaster recovery

11. checkpoint and truncate log on checkpoint