hi, just want to find out why flashcopy is so much faster than a
backup/restore method?... any directions to any helps

yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> hi, just want to find out why flashcopy is so much faster than a
> backup/restore method?... any directions to any helps

Have you read any of the manuals on the subject ? The ESS (presuming
you are using ESS) copies the data internally to another LUN, be aware
it is not an instaneous copy (contrary to what IBM may say) it is a
background process that takes place in the ESS. Backing up and
restoring from tape is only as quick as the tape drive and disk to
disk copy is a lot quicker.

Steve

hi all
i hope that you all will be remembering my previous posts regarding
ESS flash copy services.
I was trying to integrate ESS flash copy services with Oracle cold and
hot backups.
now i have acheived nearly all tasks but some questions in my mind.
BTW i have adopted the following procedure:
For Oracle cold backup for example
1.Shutdown Oracle database.
2.Some posts in same group says to sync the memory ( sync command
)before establishing flash copy pair tasks.Is this mandatory....i have
done all testings without doing sync command.
3.I did not unmount oracle filesystems or varyoffvg oravg.. is it
right to just enough to shutdown the oracle DB before establishing
flash copy pairs.
4.I establish the flash copy pair and just wait for the ESS web
interface logs to say that task is completed sucessfully.
5.start the oracle DB.
5.I made the copyvg available to either same AIX node using recreatevg
command.
6.mount all copied fs and copy the oracle data/control files to tape.
The pair is withdrawn before step 5 or step 6 and there is no problem
either.
Now when i see a previous post in same forum it says to break vg locks
using varyonvg -b command .... i do not understand why???
Is it my procedure looks ok or i need some more modifications for more
consistent backup...
please advice!!!!!
SASA

steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message news:<42862645.0401070230.6cd7da76@posting.google. com>...
> yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > hi, just want to find out why flashcopy is so much faster than a
> > backup/restore method?... any directions to any helps
>
> Have you read any of the manuals on the subject ? The ESS (presuming
> you are using ESS) copies the data internally to another LUN, be aware
> it is not an instaneous copy (contrary to what IBM may say) it is a
> background process that takes place in the ESS. Backing up and
> restoring from tape is only as quick as the tape drive and disk to
> disk copy is a lot quicker.
>
> Steve

sasa1973pk@yahoo.com (sasa queer) wrote in message news:<702930ee.0401072346.1b17c521@posting.google. com>...
> hi all
> i hope that you all will be remembering my previous posts regarding
> ESS flash copy services.
> I was trying to integrate ESS flash copy services with Oracle cold and
> hot backups.
> now i have acheived nearly all tasks but some questions in my mind.
> BTW i have adopted the following procedure:
> For Oracle cold backup for example
> 1.Shutdown Oracle database.
> 2.Some posts in same group says to sync the memory ( sync command
> )before establishing flash copy pair tasks.Is this mandatory....i have
> done all testings without doing sync command.
> 3.I did not unmount oracle filesystems or varyoffvg oravg.. is it
> right to just enough to shutdown the oracle DB before establishing
> flash copy pairs.
> 4.I establish the flash copy pair and just wait for the ESS web
> interface logs to say that task is completed sucessfully.
> 5.start the oracle DB.
> 5.I made the copyvg available to either same AIX node using recreatevg
> command.
> 6.mount all copied fs and copy the oracle data/control files to tape.
> The pair is withdrawn before step 5 or step 6 and there is no problem
> either.
> Now when i see a previous post in same forum it says to break vg locks
> using varyonvg -b command .... i do not understand why???
> Is it my procedure looks ok or i need some more modifications for more
> consistent backup...
> please advice!!!!!
> SASA
>
> steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message news:<42862645.0401070230.6cd7da76@posting.google. com>...
> > yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > > hi, just want to find out why flashcopy is so much faster than a
> > > backup/restore method?... any directions to any helps
> >
> > Have you read any of the manuals on the subject ? The ESS (presuming
> > you are using ESS) copies the data internally to another LUN, be aware
> > it is not an instaneous copy (contrary to what IBM may say) it is a
> > background process that takes place in the ESS. Backing up and
> > restoring from tape is only as quick as the tape drive and disk to
> > disk copy is a lot quicker.
> >
> > Steve

Thats an extra safe/slow way of doing it and quite lengthy depending
on the size of your db ...

No sure how Oracle will cope, but with DB2 you don't even need to stop
the database.

Just issue a write suspend on the db and flash the db volume group
which includes jfs and raw LVs, then bring it back online and varyonvg
the flashed vg, fsck the flashed pair for the jfs and bring the
flashed db online ...

We tested this to destruction (in conjunction with IBM support)
running numerous flashes at the same time having files open on the
filesystems etc ... and have been running this for over 12 months with
no problems at all, db2 / AIX cope fine ... its so robust that if it
dies half way through the flash, you just need to run it again and
works like a charm.

Hope this helps
Mark Taylor

steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message news:<42862645.0401070230.6cd7da76@posting.google. com>...
> yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > hi, just want to find out why flashcopy is so much faster than a
> > backup/restore method?... any directions to any helps
>
> Have you read any of the manuals on the subject ? The ESS (presuming
> you are using ESS) copies the data internally to another LUN, be aware
> it is not an instaneous copy (contrary to what IBM may say) it is a
> background process that takes place in the ESS. Backing up and
> restoring from tape is only as quick as the tape drive and disk to
> disk copy is a lot quicker.
>
> Steve


a LUN will be a group of disks in this case.

okay, so its a background process. then, it will take time but
definitely faster than the traditional backup/restore methods.

finally, the medium chosed will affect the speed as well. from your
last statement, its obvious that disk to disk copy is much faster than
backup/restore. therefore, the disk to disk copy is done by the
background process in ESS

am i right?

"mark taylor" <mky@talk21.com> wrote in message
news:dee0a7c3.0401080420.76d0577@posting.google.co m...
> sasa1973pk@yahoo.com (sasa queer) wrote in message
news:<702930ee.0401072346.1b17c521@posting.google. com>...
> > hi all
> > i hope that you all will be remembering my previous posts regarding
> > ESS flash copy services.
> > I was trying to integrate ESS flash copy services with Oracle cold and
> > hot backups.
> > now i have acheived nearly all tasks but some questions in my mind.
> > BTW i have adopted the following procedure:
> > For Oracle cold backup for example
> > 1.Shutdown Oracle database.
> > 2.Some posts in same group says to sync the memory ( sync command
> > )before establishing flash copy pair tasks.Is this mandatory....i have
> > done all testings without doing sync command.
> > 3.I did not unmount oracle filesystems or varyoffvg oravg.. is it
> > right to just enough to shutdown the oracle DB before establishing
> > flash copy pairs.
> > 4.I establish the flash copy pair and just wait for the ESS web
> > interface logs to say that task is completed sucessfully.
> > 5.start the oracle DB.
> > 5.I made the copyvg available to either same AIX node using recreatevg
> > command.
> > 6.mount all copied fs and copy the oracle data/control files to tape.
> > The pair is withdrawn before step 5 or step 6 and there is no problem
> > either.
> > Now when i see a previous post in same forum it says to break vg locks
> > using varyonvg -b command .... i do not understand why???
> > Is it my procedure looks ok or i need some more modifications for more
> > consistent backup...
> > please advice!!!!!
> > SASA
> >
> > steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message
news:<42862645.0401070230.6cd7da76@posting.google. com>...
> > > yls177@hotmail.com (yls177) wrote in message
news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > > > hi, just want to find out why flashcopy is so much faster than a
> > > > backup/restore method?... any directions to any helps
> > >
> > > Have you read any of the manuals on the subject ? The ESS (presuming
> > > you are using ESS) copies the data internally to another LUN, be aware
> > > it is not an instaneous copy (contrary to what IBM may say) it is a
> > > background process that takes place in the ESS. Backing up and
> > > restoring from tape is only as quick as the tape drive and disk to
> > > disk copy is a lot quicker.
> > >
> > > Steve
>
> Thats an extra safe/slow way of doing it and quite lengthy depending
> on the size of your db ...
>
> No sure how Oracle will cope, but with DB2 you don't even need to stop
> the database.
>
> Just issue a write suspend on the db and flash the db volume group
> which includes jfs and raw LVs, then bring it back online and varyonvg
> the flashed vg, fsck the flashed pair for the jfs and bring the
> flashed db online ...
>
> We tested this to destruction (in conjunction with IBM support)
> running numerous flashes at the same time having files open on the
> filesystems etc ... and have been running this for over 12 months with
> no problems at all, db2 / AIX cope fine ... its so robust that if it
> dies half way through the flash, you just need to run it again and
> works like a charm.
>
> Hope this helps
> Mark Taylor

With ORACLE the database instance should be placed into online backup mode
such that all new entries are held in the archive log. Then you would
perform the Flashcopy, which has the possibility of two modes, background
copy and non-background copy. The reason not to perform a background copy is
for backup purposes extra resource for unnecessary copying is not used and
only changed tracks from T0 (Point of initiation) are copied prior to the
target volume prior to being updated.

So once the relationship has been established (dependant on number of
relationships) between target and source, maybe 2 minutes or so, you can
place the database instance back into standard online mode.

Note for the above you will need to switch archive logging on.

Only perform background copying if you want a quick non tape restore and
only once the copy is finished.

Regards

Paresh Chudasma

Hi paresh and taylor,
Thanks for ur help!!!
But my question remains unanswered.....
Is it require to unmount filesystems in source lun Volume group or
not...
In case of online Oracle backup ( so called hot backup ) as indicated
by
Paresh , it seems nearly impossible as if i unmount filesystems of
source LUN vg
(that is Oracle vg) how the Oracle DB remain available to users during
backup...
am i right???
So please assure my conclusion to not to unmount source filesystems
before flash copy.
BUT PLEASE also see the statement from IBM about flash copy that it
copies data
bit by bit only which is present on ESS...data in filesystems cache or
DB cache will not be copied.....??????
And probably to sync changed data in fs cache the only way is to
unmount the fs.......if this is the case then ESS flash copy will not
work for Oracle or DB2 hot backups..........any ideas!!!!!!
SASA
yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401080712.58a9e9ba@posting.google. com>...
> steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message news:<42862645.0401070230.6cd7da76@posting.google. com>...
> > yls177@hotmail.com (yls177) wrote in message news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > > hi, just want to find out why flashcopy is so much faster than a
> > > backup/restore method?... any directions to any helps
> >
> > Have you read any of the manuals on the subject ? The ESS (presuming
> > you are using ESS) copies the data internally to another LUN, be aware
> > it is not an instaneous copy (contrary to what IBM may say) it is a
> > background process that takes place in the ESS. Backing up and
> > restoring from tape is only as quick as the tape drive and disk to
> > disk copy is a lot quicker.
> >
> > Steve
>
>
> a LUN will be a group of disks in this case.
>
> okay, so its a background process. then, it will take time but
> definitely faster than the traditional backup/restore methods.
>
> finally, the medium chosed will affect the speed as well. from your
> last statement, its obvious that disk to disk copy is much faster than
> backup/restore. therefore, the disk to disk copy is done by the
> background process in ESS
>
> am i right?

"sasa queer" <sasa1973pk@yahoo.com> wrote in message
news:702930ee.0401090619.3330f438@posting.google.c om...
> Hi paresh and taylor,
> Thanks for ur help!!!
> But my question remains unanswered.....
> Is it require to unmount filesystems in source lun Volume group or
> not...
> In case of online Oracle backup ( so called hot backup ) as indicated
> by
> Paresh , it seems nearly impossible as if i unmount filesystems of
> source LUN vg
> (that is Oracle vg) how the Oracle DB remain available to users during
> backup...
> am i right???
> So please assure my conclusion to not to unmount source filesystems
> before flash copy.
> BUT PLEASE also see the statement from IBM about flash copy that it
> copies data
> bit by bit only which is present on ESS...data in filesystems cache or
> DB cache will not be copied.....??????
> And probably to sync changed data in fs cache the only way is to
> unmount the fs.......if this is the case then ESS flash copy will not
> work for Oracle or DB2 hot backups..........any ideas!!!!!!
> SASA
> yls177@hotmail.com (yls177) wrote in message
news:<c06e4d68.0401080712.58a9e9ba@posting.google. com>...
> > steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message
news:<42862645.0401070230.6cd7da76@posting.google. com>...
> > > yls177@hotmail.com (yls177) wrote in message
news:<c06e4d68.0401062330.59320ffc@posting.google. com>...
> > > > hi, just want to find out why flashcopy is so much faster than a
> > > > backup/restore method?... any directions to any helps
> > >
> > > Have you read any of the manuals on the subject ? The ESS (presuming
> > > you are using ESS) copies the data internally to another LUN, be aware
> > > it is not an instaneous copy (contrary to what IBM may say) it is a
> > > background process that takes place in the ESS. Backing up and
> > > restoring from tape is only as quick as the tape drive and disk to
> > > disk copy is a lot quicker.
> > >
> > > Steve
> >
> >
> > a LUN will be a group of disks in this case.
> >
> > okay, so its a background process. then, it will take time but
> > definitely faster than the traditional backup/restore methods.
> >
> > finally, the medium chosed will affect the speed as well. from your
> > last statement, its obvious that disk to disk copy is much faster than
> > backup/restore. therefore, the disk to disk copy is done by the
> > background process in ESS
> >
> > am i right?

If you are not using a DB then yes it is advisable to sync and unmount the
filesystem prior to flashcopy. If you use a DB then placing into onlune
backup will flush database memory area. Then you can flashcopy.

It is not necessary to unmount a filesystem when using a database like
Oracle or DB2 once placed into backup mode.

Regards

On 9 Jan 2004 06:19:00 -0800, sasa1973pk@yahoo.com (sasa queer)
wrote:

>Hi paresh and taylor,
>Thanks for ur help!!!
>But my question remains unanswered.....
>Is it require to unmount filesystems in source lun Volume group or
>not...
>In case of online Oracle backup ( so called hot backup ) as indicated
>by
>Paresh , it seems nearly impossible as if i unmount filesystems of
>source LUN vg
>(that is Oracle vg) how the Oracle DB remain available to users during
>backup...
>am i right???
>So please assure my conclusion to not to unmount source filesystems
>before flash copy.
>BUT PLEASE also see the statement from IBM about flash copy that it
>copies data
>bit by bit only which is present on ESS...data in filesystems cache or
>DB cache will not be copied.....??????
>And probably to sync changed data in fs cache the only way is to
>unmount the fs.......if this is the case then ESS flash copy will not
>work for Oracle or DB2 hot backups..........any ideas!!!!!!
>SASA

Hello

Your right, anything not committed to disk from the cache will not be
flash copied. Paresh is also right, putting the likes of Oracle into
online backup mode will commit everything to disk. Other databases
also have facilities to purge any cached data as well.

Whilst it is not a requirement to unmount filesystems before
initiating a flash copy, it is the "safest" way. You do get the
benefit of not having to fsck the filesystems when re-mounted & you
can guarantee all the data has been flushed to disk.

However, it's not the end of the world if you can't/don't want to do
that. I used to use flash copy to backup a progress database which
worked like a charm without unmounting filesystems.

Paresh said earlier that it took about 2 mins to establish the flash
copy, In my (limited) experience it has never taken more than 10
seconds.

One more thing, you can use the CLI provided with your shark to
perform the flash copy operations on the host machine.

Thanks

Steven

"Steven Langdale" <steven.langdale@toyland.demon.co.uk> wrote in message
news:dmpvvv0svghpcpsprtte4t2vuvi05msktm@4ax.com...
> On 9 Jan 2004 06:19:00 -0800, sasa1973pk@yahoo.com (sasa queer)
> wrote:
>
> >Hi paresh and taylor,
> >Thanks for ur help!!!
> >But my question remains unanswered.....
> >Is it require to unmount filesystems in source lun Volume group or
> >not...
> >In case of online Oracle backup ( so called hot backup ) as indicated
> >by
> >Paresh , it seems nearly impossible as if i unmount filesystems of
> >source LUN vg
> >(that is Oracle vg) how the Oracle DB remain available to users during
> >backup...
> >am i right???
> >So please assure my conclusion to not to unmount source filesystems
> >before flash copy.
> >BUT PLEASE also see the statement from IBM about flash copy that it
> >copies data
> >bit by bit only which is present on ESS...data in filesystems cache or
> >DB cache will not be copied.....??????
> >And probably to sync changed data in fs cache the only way is to
> >unmount the fs.......if this is the case then ESS flash copy will not
> >work for Oracle or DB2 hot backups..........any ideas!!!!!!
> >SASA
>
> Hello
>
> Your right, anything not committed to disk from the cache will not be
> flash copied. Paresh is also right, putting the likes of Oracle into
> online backup mode will commit everything to disk. Other databases
> also have facilities to purge any cached data as well.
>
> Whilst it is not a requirement to unmount filesystems before
> initiating a flash copy, it is the "safest" way. You do get the
> benefit of not having to fsck the filesystems when re-mounted & you
> can guarantee all the data has been flushed to disk.
>
> However, it's not the end of the world if you can't/don't want to do
> that. I used to use flash copy to backup a progress database which
> worked like a charm without unmounting filesystems.
>
> Paresh said earlier that it took about 2 mins to establish the flash
> copy, In my (limited) experience it has never taken more than 10
> seconds.
>
> One more thing, you can use the CLI provided with your shark to
> perform the flash copy operations on the host machine.
>
> Thanks
>
> Steven
>

Steve,

I put down two minutes because the time is dependent on the number of
relationships the need to be establishe and the load on the Shark. I know a
couple of minutes is way off but we don't know the relationships and stress
factor.

To use the CLI you will need a user and password within copy services
adminstration and JAVA 1.1.8 or 1.3 depending on versions.

Regards.