Database backup without using 'backup database'

[Bernard Dhooghe]

Suppose a running database (DB2 UDB) with all containers being SMS
based (file system as JFS or JFS2 on AIX).

How to make a backup (online) by doing a file system backup and not a
database backup?

Could this be done with:

db2 set write suspend for db

backup file system data used by db2 data and logs

db2 set write resume for db

with the purpose of doing:

restore of db2 file system data on another system
startup and usage of database.

The questions are:

- will the restored database be consistent after start/restart

- what additional resources are taken between suspend write and
resume write (a file system backup takes more time that just splitting
off a snapshot in a storage subsystem).

Bernard Dhooghe

[Pierre Saint-Jacques]

What you are attemting is pzartially correct.
Putting your running db in write suspend is the first step.
Using file system copy to the target is the next. It depends on what how
you plan to use the db on the target: standby, snapshot or mirror.
You will have more or less to copy.
Once copied the write resume is the proper way.

On the target, you have to ensure that either ALL of the following are true
and the same:
Instance name, db bname, dbpath,logpath,archivelogpath,all container paths.
Then you can use the:
db2inidb command with the proper parm.

If one or some or all of the above change, you will have to read the docs
and learn to use the relocatedb command to generate the specific file that
will tell what to relocate.

HTH, Pierre.

--
Pierre Saint-Jacques
SES Consultants Inc.
514-737-4515
"Bernard Dhooghe" <dh******@yahoo.com> a écrit dans le message de news:
11*********************@j55g2000cwa.googlegroups.c om...

Suppose a running database (DB2 UDB) with all containers being SMS
based (file system as JFS or JFS2 on AIX).

How to make a backup (online) by doing a file system backup and not a
database backup?

Could this be done with:

db2 set write suspend for db

backup file system data used by db2 data and logs

db2 set write resume for db

with the purpose of doing:

restore of db2 file system data on another system
startup and usage of database.

The questions are:

- will the restored database be consistent after start/restart

- what additional resources are taken between suspend write and
resume write (a file system backup takes more time that just splitting
off a snapshot in a storage subsystem).

Bernard Dhooghe

[hikums]

Pierre,

Can you please elaborate the following step for any product that you
can(AIX or Linux preferred):

Pierre: Using file system copy to the target is the next. It depends
on what how
you plan to use the db on the target: standby, snapshot or mirror.
You will have more or less to copy.
Kums: say I want to do a mirror, that way it will be a backup, and I
have to rollforward. How do I actually do the filesystem copy? The unix
or vendor product based steps are needed to get an understanding.

This is not elaborated in any manuals as it may depend on the vendor
product, but I am curious to know the exact steps.

Thanks very much in advance.

Kums

[Pierre Saint-Jacques]

The following link will give you quite a lot: Just to let you know, a lot
of info on DB2 is in the Google db.
Going to Google search: db2 and "split mirror" produced that as the first
hit.
http://www-128.ibm.com/developerwork...0204quazi.html

As you see it is V7 but it applies to V8 also.

If you initdb as a mirror, you can do maintenance and backup on the copy and
then restablish the mirrors in a reverse way.

Now for the file system copies, it depends on what disk hardware you are
using. It can be as crude as doing an OS file copy to tape and copying to
disk on the target and as fancy as using Flash Copy services on an IBM ESS
box.

If you use DB2 Backup commands, it could be a little more interesting.

Let me know if you need more, Pierre.

--
Pierre Saint-Jacques
SES Consultants Inc.
514-737-4515
<hi****@gmail.com> a écrit dans le message de news:
11**********************@j55g2000cwa.googlegroups. com...

Pierre,

Can you please elaborate the following step for any product that you
can(AIX or Linux preferred):

Pierre: Using file system copy to the target is the next. It depends
on what how
you plan to use the db on the target: standby, snapshot or mirror.
You will have more or less to copy.
Kums: say I want to do a mirror, that way it will be a backup, and I
have to rollforward. How do I actually do the filesystem copy? The unix
or vendor product based steps are needed to get an understanding.

This is not elaborated in any manuals as it may depend on the vendor
product, but I am curious to know the exact steps.

Thanks very much in advance.

Kums