RE: Falconstor/IPStor with Clariion, EMC Timefinder with Symmetrix

  • From: "Lim, Binley" <Binley.Lim@xxxxxxxxxx>
  • To: "'oracle-l@xxxxxxxxxxxxx'" <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 25 Mar 2004 12:48:10 +1200

Netapps works in a similar fashion.

What they forgot to mention is the following side effect. Because you are
writing to new blocks on updates, you would consume more and more space and
will eventually run out of space if you let that continue.

From an Oracle perspective, it is only doing updates so it is not requesting
more space. So it doesn't handle very gracefully the situation where you do
run out of space. First the instance crashes. Then it refuses to start
because it says some files are not available. Then you look into the alert
log and see DBWR had decided to offline the datafiles. Simple fix - just
online the datafiles mentioned in the alert log. 
But the online command fails with the error that those files are *not*
offline. Then you check the spelling to make sure you have got the right
filenames mentioned in the alert log - can't really go wrong with
cut-and-paste, but you check anyway. Nope, no typos there. 

Start again. Verify the space problem has been fixed (by deleting
snapshots). Shutdown the instance, and attempt startup. Nope, same problem.
Says files are offline. Try onlining the files. Nope, says files are not
offline. Geez...

Finally, after dumping some controlfiles and file_headers and v$views in an
un-opened state, I noticed there was inconsistency in that the datafiles
were flagged as offline and online in different places. Not sure how I came
to think of the solution, but it was to explicitly offline everything, and
then online them. That worked, and the instance went through the process of
recovery and started. Whew!

> -----Original Message-----
> From: Yechiel Adar [SMTP:adar76@xxxxxxxxxxxx]
> Sent: Thursday, March 25, 2004 4:33 AM
> To:   oracle-l@xxxxxxxxxxxxx
> Subject:      Re: Falconstor/IPStor with Clariion, EMC Timefinder with
> Symmetrix
> 
> Hello Lex
> 
> I think that you got it wrong. The storage system does not write the block
> twice.
> Instead it will write the new block to another place on the disk and will
> update the pointers only. The old block stays in place and can not be
> overwritten as it is marked as used by the snapshot.
> 
> Yechiel Adar
> Mehish
> ----- Original Message -----
> From: "Lex de Haan" <lex.de.haan@xxxxxxxxxxxxxx>
> To: <oracle-l@xxxxxxxxxxxxx>
> Sent: Wednesday, March 24, 2004 3:26 PM
> Subject: RE: Falconstor/IPStor with Clariion, EMC Timefinder with
> Symmetrix
> 
> 
> > Mike, Carel-Jan,
> >
> > this sounds like a pretty expensive solution (and I am not talking money
> > here) -- just imagine how many I/O activity you typically have against
> an
> > online Oracle database... this means that after the "snapshot point in
> time"
> > you basically start writing every block twice -- once to the database,
> and
> > once to the snapshot.
> >
> > well, I can tell you, the Oracle mechanism (activated by putting files
> in
> > backup mode) is definitely cheaper, because it is a more intelligent
> > algorithm; it only writes full block images to the redo log when needed
> for
> > recoverability, and definitely NOT for every change ...
> >
> > cheers,
> > Lex.
> >
> 

This communication is confidential and may contain privileged material.
If you are not the intended recipient you must not use, disclose, copy or 
retain it.
If you have received it in error please immediately notify me by return email
and delete the emails.
Thank you.
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Other related posts: