After Image prep steps for Dump & Load

R

RayL

New Member
9.1D09 (I know...) on HPUX 11.11, 35GB database

I'm planning a D&L and would like to know if there are any special considerations for After Imaging. Is there any reason / advantage to run "aimage end" in this (very rough) scenario? If so, why & when?

(Save current)
Shutdown
Offline backup
Save the ai files
prorest to temp location as validation

(Dump)
dump sequences _seqvals.d
dump the tables (example: $DLC/bin/proutil database -C DUMP table /temp/bindump/ -RO)

(Load)
build the empty database
load the incremental df
backup
load the data (example: $DLC/bin/proutil database -C LOAD table.db -1 -i -yy 1950)
load the sequences
backup
idxbuild

(Go live)
backup
aimage begin

Thanks,
Ray
 
No, there is no point to ending AI. If you abort the D&L and end up continuing to use the old system you will want AI to still be running.
 
Why would you use a single user database connection for the load of the data.
Just start up de database with a resonable large -B and start the load script for all tables.

A simple aproach would be:

Dump --> start database with high -B:
-Dump _user, sequence current values, data definitions
-Dump data (binary)

Use the performance measurements and analysis you made on the data before the dump to create a new df file. (This is an excellent opportunity to reorganise your database to avoid the reason to do a reload).

Load:
I would skip the extra backup step you have between loading df and loading the data. This is just overkill. If the load of data fails it is a matter of seconds to make a new database and load the df again, so nothing to gain there.

-Create new database with the right blocksize (4 or 8K depending on OS in V9) and the modified df.
-Start database with sufficient -B -i and -G 0
-Load df, _user, sequence current values
-Load data
-Rebuild indexes
-Backup the database
-Enable AI

I don't see why you would need an incremental df for a dump/load of the database. So I assume you mean df file (without incremental).

I think there should be 1 step added:
Step 1: Upgrade to OE10.2A
Step2: all of the above with taking advantage of nice stuff like typeII storage area's

And if you find any troubles with this, I am sure Tom is more then willing to guide you in making the right decissions.


Regards,
Casper.
 
Tom, good point about not ending ai - keeps my options open if something goes wrong with the D&L.

Casper, we are a QAD shop, running an older version (eB SP4). So v10 is not an approved version for this implementation. As far as the D&L process goes, I've done some benchmarks with the database up (various -B sizes) and down and find that parallel dump (typically 2) and serial load works best on this box.

Thanks for the responses,
Ray
 
It's a hard sell. Our environment is rock solid and we don't have any major enhancements planned. Based on my benchmarks I can't justify going to 9.1E04 on D&L performance or available features. The hope is to move to V10 as part of a QAD upgrade. Still waiting for the green light on that one...
 
You must log in or register to reply here.
Back
Top