Re: IB2007 journals and archiving question

Michael Danninger wrote:

> Is it possible to produce "corrupt" journal files by using
cached
> writes to the journal files or by maybe a network connection loss
> while archiving the journals over a network? If so, what happens if
> one recovers from the corrupt journal files? Is the recovered
> database corrupt or will ib notice the corrupt files and gbak stop
> recovering?

InterBase will not cache writes to the journal files under any
circumstances. The big risk from caching comes from putting the
journals on a drive with a caching controller.

> 
> This questions came to my mind, because i have a server with a raid
> system, and no possibility to put the journal file onto a drive,
> which is not cached. Since there are no performance problems, it
> would be fine for me to put the journals on a logical drive on the
> raid. I plan to archive the journals periodicaly to a backup server
> over the network. Are there any advices to use journals with raid
> systems?

The journal files are sequential. When you recover IB reads the journal
file starting at the beginning and reads until it cannot read anymore
so my understanding is that you will recover what actually got written
to the journal.

-- 

Michael Danninger wrote:

> impossible to corrupt

I will never use the phrase "impossible to corrupt" for any file on
any
disk on any computer. The type of file that is least susceptible to
corruption is a sequential file and the journal files are sequential.
It is very very very very unlikely that a journal file will become
corrupt.

-- 

Is it possible to produce "corrupt" journal files by using cached
writes to 
the journal files or by maybe a network connection loss while archiving the 
journals over a network? If so, what happens if one recovers from the 
corrupt journal files? Is the recovered database corrupt or will ib notice 
the corrupt files and gbak stop recovering?

This questions came to my mind, because i have a server with a raid system, 
and no possibility to put the journal file onto a drive, which is not 
cached. Since there are no performance problems, it would be fine for me to 
put the journals on a logical drive on the raid. I plan to archive the 
journals periodicaly to a backup server over the network. Are there any 
advices to use journals with raid systems?

Thanks in advance,
Michael 

> InterBase will not cache writes to the journal files under any
> circumstances. The big risk from caching comes from putting the
> journals on a drive with a caching controller.
>
> The journal files are sequential. When you recover IB reads the journal
> file starting at the beginning and reads until it cannot read anymore
> so my understanding is that you will recover what actually got written
> to the journal.

So is it right, that the worst thing that can happen using a caching 
controller is to loose the transactions in the cache, but that it is 
impossible to corrupt the journal file, to which the cache should have been 
written? 

> I will never use the phrase "impossible to corrupt" for any file
on any
> disk on any computer. The type of file that is least susceptible to
> corruption is a sequential file and the journal files are sequential.
> It is very very very very unlikely that a journal file will become
> corrupt.

Thanks a lot for your answers, Michael