After I remove the online fuzzy tag of the datafile, can it be safe on standby?
AnySQL (d.c.b.a)
I developed a database file tool, when database is open, I do not
put table into "begin backup" mode, just clear the fuzzy tag on data
file header, and make sure that no fractured block in the copied
datafile, seems like a cold copy, but SCN on some block may be newer
than the datafile header's.
Is it safe for this file to be used in standby?
Mladen Gogala
> Is it safe for this file to be used in standby?
Probably not. You are trying to develop an alternative to RMAN,
at a lower cost? Anybody using your tool is likely to immediately lose
database support. I very much doubt that Oracle Corp. would support
the database in which file headers were changed by anything except
DBWR or CKPT.
sybr...@yahoo.com
Whether or not it is safe is irrelevant. What is relevant,is that
Oracle obviously won't support your little tricks, as you are violating
your license by reverse engineering Oracle.
--
Sybrand Bakker,
Senior Oracle DBA
fitzj...@cox.net
Comments embedded.
AnySQL (d.c.b.a) wrote:
> Hi all:
>
> I developed a database file tool,
Really. By definition a tool is a device designed for the reliable
performance of a task, and it occurs to me you've not met that
criteria.
> when database is open, I do not put table into "begin backup" mode,
I see, so this 'tool' is your idea of a backup utility ...
> just clear the fuzzy tag on data file header, and make sure that no fractured block in the copied datafile,
You dramatically alter the datafile header by clearing the 'fuzzy' tag,
without verifying any further information in the file, such as SCN
consistency. This sounds more like a data shredder.
> seems like a cold copy, but SCN on some block may be newer than the datafile header's.
And with enough body work, suspension work, powertrain work, upholstery
and paint a Yugo can seem like a Rolls-Royce, although it never will
actually BE a Rolls-Royce. However, you've skipped straight to the
paint work, and foregone any of the other tasks necessary to convert
that Yugo to a possible pseudo-Rolls, and yet you're trying to sell it
as though you have. Is this file copy made from a bonafide backup?
No, by your own admission. Yet you want other people to accept your
poor excuse for a product and go merrily on their way, trusting your
modified files to contain their production data. I believe you need to
return to reality (you remember, the place where people display a solid
grasp of conceptual accuity), and possibly undergo some serious
therapy, as you're obviously delusional.
> Is it safe for this file to be used in standby?
Personally I wouldn't use the file as a doorstop, much less as a
component of a standby database. A standby database is intended to
shorten recovery time in the event of a disaster, not create the
disaster by using files modified by a 'tool' that only changes the
headers without checking for other possible problems. Certainly it's
possible to drive screws with a sledgehammer, but that isn't the tool
specifically designed for the operation. It's also possible to use a
boat anchor as a toothpick, presuming you can hoist it to your mouth
and force it between your teeth; of course, that isn't very PROBABLE,
for good reason. It isn't the proper implement for the task. Neither
is your 'tool' the proper implement to create standby database files.
And I know of no one who would jeopardise their service agreement, or
their data, with your six-lane superhighway to ruin.
What, exactly, are you trying to prove with this fiasco you call a
'tool', prove yourself smarter than the Oracle developers? You are
sorely mistaken if that is your aim. Are you trying to write a tool
which performs the same function as RMAN, only without Oracle's 'high
licencing costs'? You've missed the mark, sad to say, as you've done
nothing except prove you know so little about Oracle as to be
considered dangerous. One does not re-write datafile headers, nor does
one try to prove himself intelligent by basically writing an automatic
hex editor to modify data in a way Oracle did not intend. Such antics
only create problems, not solutions. Altering a datafile header
without the proper tools is asking for trouble, and possibly a lawsuit.
If you market this wonderous software package, purchase PLENTY of
liability insurance. Trust me, you'll be makkng claims on the policy
until the carrier cancels it, as an Oracle database isn't safe when
this 'tool' is used.
David Fitzjarrell
AnySQL (d.c.b.a)
put tablespace in hot backup mode or rman when building standby, I read
the data file header first absolutely, and then read the others block,
as a matter of fact, I did a simple test on my small database installed
on my notebook with my tool, It works.
One thing not sure is that how oracle determine the start SCN of
recovery for a datafile? Accord SCN in the data file headers or others
SCN in the data file header?
Thanks.
AnySQL (d.c.b.a)
database, If I put database into begin backup mode, too more extra log
generated, and if using rman, I must copy to a local directory, and
then transfer to remote host, thus make the database host overload on
IO.
Thanks.
DA Morgan
Which part of violating Oracle's license for their product did you not
get from every previous response you have received?
Strikes me the only one interested in doing business with you may well
be an attorney working for Safra Catz.
--
Daniel A. Morgan
http://www.psoug.org
damo...@x.washington.edu
(replace x with u to respond)
gazzag
follow Oracle's documented and, more importantly, supported procedure.
Chuck
news:1120571223.6...@g49g2000cwa.googlegroups.com:
I'm not sure I see what the benefit of suh a tool would be, even if it
worked (which I don't think it would). It sounds like you're trying to
save a very small amount of processing time at the risk of corrupting your
data. It also sounds like a database Darwin Award in the making if you ask
me.
Either use rman, or begin/end backup and copy/ftp your files. It's
platform independent, safe, and reliable.
fitzj...@cox.net
Comments embedded.
AnySQL (d.c.b.a) wrote:
> put tablespace in hot backup mode or rman when building standby,
You're BUILDING a standby database by editing the datafile header using
a 'tool' (and I use the term LOOSELY) which Oracle has neither written
nor approved. This is a guarantee of a lawsuit from Oracle if I ever
saw one.
> I read
> the data file header first absolutely, and then read the others block,
> as a matter of fact, I did a simple test on my small database installed
> on my notebook with my tool, It works.
It works until you have a crash and can't recover due to your file
header manipulation.
>
> One thing not sure is that how oracle determine the start SCN of
> recovery for a datafile? Accord SCN in the data file headers or others
> SCN in the data file header?
>
Again, I state you have only proven yourself less than qualified to
touch an Oracle database, much less write 'faulty' tools to perform
unspeakable acts against the datafiles. Taking shortcuts inevitably
leads to frustration by circumventing known, reliable, standard
procedures to perform the same act. Hot backups can be used to create
standby databases; your 'home-brewed' method of subverting that
procedure is a sure route to failure, as your incomplete changes will
result in a lost datafile, or a lost database due to file header
corruption creating an inability to recover the files. This equals
lost data, lost time, lost revenue and excessive expense for any
potential customer who may feel brave enough to use your 'product',
along with lawsuits against you by the customer, and by Oracle
Corporation for attempted reverse-engineering of the Oracle software.
> Thanks.
>
You're welcome. Now stop this nonsense and use the tools Oracle
provides.
I've visited your website, even though I don't read Chinese. The fact
you claim yourself as an Oracle Certified Professional is disturbing,
to say the least, as one would expect an OCP to respect the software
from Oracle corporation, not try to pull an 'end run' around it to cut
corners.
David Fitzjarrell
AnySQL (d.c.b.a)
fun, as a matter of fact, we did not use any tool until oracle say it's
safe.
Try to know more of how oracle perform recovery.
Thanks, everybody.
DA Morgan
Mmy impression is that "for fun":
1. You voided your employer's support agreement
2. You violated your employer's license agreement
3. You put your employer at risk of legal action (however unlikely)
Not my idea of a day's fun.
And were you my employee not my idea of someone I'd call a "keeper."
To use the previous race car analogy ... you learned how to use the
throttle. Too bad you didn't also read the speed limit signs before
you put the pedal to the metal.
AnySQL (d.c.b.a)
it will be better for us to build standby on very busy database system.
sybr...@yahoo.com
Search for a different profession, or your employer will sooner or
later regret he ever hired you (after he went bankrupt because of your
actions)
--
Sybrand Bakker
Senior Oracle DBA
Holger Baer
> I think oracle can integrite rman copy feature into rsync or scp etc,
> it will be better for us to build standby on very busy database system.
>
Duh?
Using rman you can set the maximum rate at which each channel may read,
reducing the impact on your database. It does not induce a higher redolog
creation as using begin backup/end backup and it takes care of block
corruption etc. You can use remote storage as target, reliefing the impact
on your local IO-Subsystem.
Now, what where you talking about?
Regards
Holger
AnySQL (d.c.b.a)
standby, just test and test on my personal computer.
How quest build your trust? It's very tough.
But now, I think you trust quest's software.
Holger Baer
What has quest to do with this thread? RMAN is an Oracle developed,
Oracle supplied and Oracle supported software. Quest is a third party
vendor that I used to use. And believe me: I never trusted them.
And I don't trust RMAN either - I did test it and the results where
repeatedly predictable. But even with RMAN, which comes very near
to be trustworthy, you can easily burn yourself if something unexpected
enters the picture and you're back to having to know what's going on
and how to recover from that.
Do I rely on RMAN? Yes.
Do I trust it? To some extent.
Who do I have to trust? My brain, and even that fails me frequently!
Regards
Hooger
Joel Garry
>standby, just test and test on my personal computer
Well, you didn't test it enough. Look at
http://download-west.oracle.com/docs/cd/B19306_01/server.102/b14220/backrec.htm#i1005950
and search for the word "fractured."
You see, what you are doing is turning off the switch that says "hey
oracle, we didn't use RMAN to back this up, so you need to recover and
make the data consistent again." This is guaranteeing that you will
screw things up. If you can't find something wrong, your testing is
inadequate. Of course, most testing usually is. You need to be
jamming enough transactions in while copying the data file that you
catch the situation where an OS block has been updated before your copy
is reading it, but the next OS block (which is part of the same Oracle
block) hasn't been updated yet. And there's no way to force it since
you don't have control of the database writer. And it is very likely
on a busy database.
You need to understand the basic concepts before you write any tools,
or anything else.
So you are saying your database is too busy to backup? Yikes. Get a
competent DBA and fix your recovery procedures before you bother with a
standby . http://www.dizwell.com/html/backup.html
jg
--
@home.com is bogus.
Bleeding edge:
http://www.signonsandiego.com/uniontrib/20050705/news_1b5hydrogen.html
AnySQL (d.c.b.a)
I think the CKPT process taged the data file header after DBWn
sucessfully write the block, for example, at SCN1 DBWR performed
writing, and the CKPT taged SCN1 to all data file header, if I read
data file header first, when oracle want to recover data file, it will
start on SCN1, if I read data file header before CKPT, then it will
start from previous SCN, so no log will be bypassed in this case.
I dump the data file header copied from rman, the online fuzzy tab is
removed, and some blocks' SCN may be larger than the data file
header's, just the same case.
What I finally want is to integrete the rman copy with scp or rsync.
Mladen Gogala
> I think oracle can integrite rman copy feature into rsync or scp etc,
> it will be better for us to build standby on very busy database system.
Yeah. Oracle can probably integrate rman into fdisk as well but I don't
really see the use for it. What are you trying to do? To build standby?
I did it many times, without ever shutting the database down.
It's done from RMAN, without any problems. The company's database is RAC
with 2 standby databases. Last weekend, there was a major SAN upgrade
and I had to switch to standby, re-create primary database and rebuild
standby. Total downtime: 0 minutes.
chao_ping
How can you manage to switch to standby without shutdown your
application? Your application did the transparent failover from primary
to standby?
If you don't mind, can you tell us your idear about how to reduce the
downtime to 0?
Build standby without shutdown primary is a everyone know stuff now:)
With Rman to build standby, your do need a shared tape library, backup
to tape, and restore from tape from the disk.
Thanks
DA Morgan
> With Rman to build standby, your do need a shared tape library, backup
> to tape, and restore from tape from the disk.
>
> Thanks
No tape is required: Ever.
chao_ping
database from one host to another?
Can you share your idea about it?
Thanks
DA Morgan
Care to tell us what operating system?
What version and edition of Oracle?
The number and size of the datafiles?
The time alotted to the process?
Tolerance for downtime?
Service level agreement with customers?
Or should we just assume that all databases are identical? ;-)
chao_ping
Does the questions you asked relate to my question?
No downtime is what we are talking about, of course. Why do we need
to shutdown databasee to build a standby/backup a database?
Don't tell me you use storage software image copy, or use NFS to do
rman copy.:)
Thanks
chao_ping
Joel Garry
I have a situation where the only link between the standby and the
primary is a network link that is far too small relative to the size of
the db to copy or share anything. So I merely copy the backup to a
tape and send someone there on a plane. The RMAN standby storage is
made to look like the primary storage through unix links, that is the
big trick to not having actual shared storage.
Sometimes I have to play funny games with crosschecking so RMAN doesn't
try to apply newer backup files that have been created on the primary
while the tape is waiting for the airplane (I'm not using a catalog).
http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96629/dbresource.htm#1019838
shows one way you can switch back and forth. I would guess Mladen had
coordinated a time to deal with any app issues, I would hope no one
would try to schedule a SAN upgrade during a busy time or when someone
is running long transactions.
jg
--
@home.com is bogus.
http://www.signonsandiego.com/uniontrib/20050707/news_lz1c07offer.html
chao_ping
When it is really necessary copy big datafiles/archivelogs through
slow network, you may try rsync -z flag. It works great when you have
relatively slow link compared to your datafile/archivelog size. It does
greatly improve speed when link slow.
And for rman restore, I simply do a full backup, and then backup
controlfile , as the controlfile now have the information about the
backup done, when restore to new box, I restore controlfile, and use
the controlfile to restore the database. (we backup to diskk and then
tar to tape that time.).