need help 
Author Message
 need help

We have a huge database running on production box and we use another box for
reporting purpose.
For this we move checkppoint from production box to report box and refresh
it with new data to run reports.

Now our system admin guys have decided to do the other way.

they want to copy data files from production data location to a NFS mounted
system (of report box) on a continous basis and want to stop at a certain
point and unmount the NFS partiton and want to use those datafile for
reporting database.

Now my question is

How can we ensure consistency of the data files?

At what point shall we stop copying the files so that all which has copied
is consistent.

is someone has some idea or might have implemented it, let me know.

Thanks in advance,

Larry

______________________________________________________
Get Your Private, Free Email at http://www.***.com/



Mon, 25 Mar 2002 03:00:00 GMT
 need help



Quote:
> [ ... DB replication ... ]
> Now our system admin guys have decided to do the other way.

> they want to copy data files from production data location to a NFS
mounted
> system (of report box) on a continous basis and want to stop at a
certain
> point and unmount the NFS partiton and want to use those datafile for
> reporting database.

> Now my question is

> How can we ensure consistency of the data files?

This won't work - you will get data inconsistencies on your reporting
Ingres system. If you want to refresh the database on that system
and use it at the same time, you may be able to use replicator (I know
nothing of Replicator, however, & NFS??!?! sniff... sniff....) - if
you just want a refresh of the database on your reporting system
every so often, then at refresh time:

i   ) Stop your production Ingres system DBMS(s).
ii  ) Un-mount the production filesystem(s) that contain your
      production data areas.
iii ) Mount the production filesystem(s) again.
iv  ) Stop your reporting Ingres DBMS(s).
v   ) Copy anything changed since last time to your reporting
      Ingres system.
vi  ) Re-start your production and reporting Ingres DBMS(s).

However, this plan may defeat the object of the exercise, whatever
that be.

Quote:
> is someone has some idea or might have implemented it, let me know.

Maybe..... what kit do you run these systems on?

Richard.
--

if you will be so kind.

Sent via Deja.com http://www.deja.com/
Before you buy.



Mon, 25 Mar 2002 03:00:00 GMT
 need help
This is very risky and chances are it won't work.

Have you considered copying closed journal files over
to the reporting box and applying them to the reporting
database to keep them near consistent ?

If
1- Both database are exactly the same at the start.
2- Journal records use table names and not reltids.

It might work.
I never tried it.

Andre

Quote:

>We have a huge database running on production box and we use another box
for
>reporting purpose.
>For this we move checkppoint from production box to report box and refresh
>it with new data to run reports.

>Now our system admin guys have decided to do the other way.

>they want to copy data files from production data location to a NFS mounted
>system (of report box) on a continous basis and want to stop at a certain
>point and unmount the NFS partiton and want to use those datafile for
>reporting database.

>Now my question is

>How can we ensure consistency of the data files?

>At what point shall we stop copying the files so that all which has copied
>is consistent.

>is someone has some idea or might have implemented it, let me know.

>Thanks in advance,

>Larry

>______________________________________________________
>Get Your Private, Free Email at http://www.hotmail.com



Mon, 25 Mar 2002 03:00:00 GMT
 need help

Quote:
>We have a huge database running on production box and we use another box
for
>reporting purpose.
>For this we move checkppoint from production box to report box and refresh
>it with new data to run reports.

>Now our system admin guys have decided to do the other way.

>they want to copy data files from production data location to a NFS mounted
>system (of report box) on a continous basis and want to stop at a certain
>point and unmount the NFS partiton and want to use those datafile for
>reporting database.

>Now my question is

>How can we ensure consistency of the data files?

>At what point shall we stop copying the files so that all which has copied
>is consistent.

>is someone has some idea or might have implemented it, let me know.

They just can't do this.  It won't come close to working.

You recognize the need for a consistent image of the database.  You can
force a consistency point by executing trace point DM1305, but that is not
enough.  In the first place you have to have some way to stall the server
while they copy all the files--which will take a long time.  And even if you
can stall it, the consistency point will post modified pages for uncommitted
transactions, even for transactions that may get rolled back, so they will
get a corrupted database.  It is just a very {*filter*}idea.  Don't do it.

Either continue to do what you are doing; investigate using replication; or
do something clever with mirrored file systems and checkpoints (I think both
Mike Leo and Karl Schendel have described this in detail several times).

Roy Hann
Rational Commerce Ltd.
(Formerly BKB Client/Server (UK) Ltd.)



Mon, 25 Mar 2002 03:00:00 GMT
 need help

Quote:
>...they want to copy data files from production data location to a NFS mounted
>system (of report box) on a continous basis and want to stop at a certain
>point and unmount the NFS partiton and want to use those datafile for
>reporting database.

>Now my question is

>How can we ensure consistency of the data files?

This is possible but nontrivial.  I can give you a rough outline, but I
have to start by suggesting that you continue doing just what you are doing.

You can snapshot a database by messing with the checkpoint template such
that instead of doing a tar to tape or disk, you instead copy the data
files to your target disk area.  When this pseudo-checkpoint is done you
take the dump files created as part of the checkpoint and copy them to the
target area.  Place all this stuff in the proper Ingres database locations
on the target side and run a rollforward -- sort of.  You have to jigger
the target's checkpoint template so that it doesn't delete the data files
you just moved in.  (Or you can arrange for the rollforward to do the
moving of the data).  There are other things to fix too, such as the
dbid and locaton list in the target's iidbdb.

As you can see, this is essentially what you do now.  The only reason that
I can see for trying to make it more complicated is if you have a fancy disk
subsystem that can do the actual disk-to-disk copy extremely quickly at
the disk array level.  (E.g.  the EMC Symmetrix's BCV feature.)
If you don't have a low level disk-to-disk capabiility, I think you are
wasting your time.  Just do the checkpoint, move over the checkpoint's tar
file, and rollforward it normally.

Karl R. Schendel, Jr.

Ingres and Unix Expertise

Past President, North American Ingres Users Association



Mon, 25 Mar 2002 03:00:00 GMT
 need help

Larry Green a crit dans le message

Quote:
>We have a huge database running on production box and we use another box
for
>reporting purpose.
>For this we move checkppoint from production box to report box and refresh
>it with new data to run reports.

Replicator could probably feet your need. Your second database can be used
to receive data, only. This, keeping consistancy and real-time.

Regards

C.Br.



Mon, 25 Mar 2002 03:00:00 GMT
 need help
We have a similar setup with a huge production database and a report
(well sort of) system.

I've seen a few answers suggesting the possible use of Replicator. If
you are talking about 6.4 then don't use replicator , especially if you
have a very high throughput. Ingres II replication is however much
better and much improved over 6.4.

Due to the problems with replicator under 6.4 we use a C program that
converts the output from auditdb into sql which is then replayed
straight into the report database. It handles much bigger volumes of
throughput than replicator.

However I stand by what the other folks have said , that the easiest way
is to refresh from checkpoint like you're doing already.

Piers Collins



Quote:
>We have a huge database running on production box and we use another box for
>reporting purpose.
>For this we move checkppoint from production box to report box and refresh
>it with new data to run reports.

>Now our system admin guys have decided to do the other way.

>they want to copy data files from production data location to a NFS mounted
>system (of report box) on a continous basis and want to stop at a certain
>point and unmount the NFS partiton and want to use those datafile for
>reporting database.

>Now my question is

>How can we ensure consistency of the data files?

>At what point shall we stop copying the files so that all which has copied
>is consistent.

>is someone has some idea or might have implemented it, let me know.

>Thanks in advance,

>Larry

>______________________________________________________
>Get Your Private, Free Email at http://www.hotmail.com

--
Piers T Collins


Tue, 26 Mar 2002 03:00:00 GMT
 
 [ 7 post ] 

 Relevant Pages 

1. help,help,help,help,help,help,help,help,help,help,help,help,help,

2. Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,Help,

3. SQL To ORACLE Help Help Help Help Help

4. query:HELP! HELP HELP HELP HELP

5. {Need Help Big Project Need Help}

6. Rookie needs help clearing the Clipper hurdle (DBase help needed)

7. HELP: Need help building a Decisoin tree help proggy

8. Help, Help, I need Help...!

9. Help, Help, Help I need a template for invoicing, orders and customers

10. HELP HELP HELP HELP

11. HELP HELP HELP HELP!

12. need help implementing instantDB, HELP Please!!!


 
Powered by phpBB® Forum Software