RE: standby instance, log recovery

Hello Raimund,

>first remark: when the kernel fails to start (into admin-mode) due to 
>the disk being full, you cant tell because knldiag.err is empty 
>(probably because the disk is full :).

The file knldiag is currently your only hope in such a situation.

>> may i ask what's so dangerous about going online in the standby 
>> instance? is it that logs are being written? but the 'clear 
>> log' should
>> help then?

Log is created during going online and a new Savepoint is written to the
data volumes at the end of it. So if you start into mode ONLINE between
restoring data and restoring the first log a 'clear log' does not help.
If you do start into mode ONLINE only before you restore data and not
between data and log restores a 'clear log' is a viable solution.

>>is it ok for the two systems to have different db-ids?

That is the way it should be.

>(it seems that the standby db has no own db-id. but how could i tell?)

As DB-IDs were invented to prevent mixing backups from different
databases, the new db gets it's new ID only when the first backup is
executed. db_restartinfo can then again be used to determine the ID.

>and i have a few more questions: (please direct me to any 
>documentation 
>that may contain the answers. would be nice if the docs would 
>be accurate :)
>
>1) am i supposed to never bring the standby-instance up unless 
>i really 
>want to replace the original instance?

Yes.

>when i bring it up once, will i 
>be able to import more log backups from the exporting instance?

Not without restoring a data backup first. To check what is in the
standby database, you could make a backup of the standby database though
and restore it into a third database.

>1a) if i am forbidden to ever bring up the standby instance, can i 
>"lock" db_online in any way so that it doesnt happen accidentally?

Not that I know.

>2) for a continuous standby instance i have to decide wether i should 
>use some script to perform log backups at regular intervals 
>(and import 
>them) or if i should use autolog. is there any advice or practical 
>experience?

Currently Autolog does not give you the opportunity to backup log after
a certain amount of time (regardless of the amount of unsaved log).

>are there ready-made scripts (i noticed some ugly 
>behaviour 
>of the exporting instance when the log is empty).

I don't know if there is a script out there. Maybe you find one in the
list archives. There is non delivered with MaxDB.

>3) can/should i mix automatic and manual log backups?

You can mix both types of log backup. The difference is in creation not
in the resulting log backups.

>4) does it feel different to import automatic logs? does it work the 
>same way?

It works the same way.

>5) does the standby instance produce logs while importing the log 
>backups of the original instance?

No.

Best regards,
Tilo Heinrich
SAP Labs Berlin

--
MaxDB Discussion Mailing List
For list archives: http://lists.mysql.com/maxdb
To unsubscribe:    http://lists.mysql.com/maxdb?unsub=gcdm-maxdb@xxxxxxxxxxx