Upgradation & Migration

narasimha · Post by **narasimha** » Fri Jun 24, 2005 3:57 pm

Though there may be different steps followed during the Upgradation and Migration of Datastage between different Versions, was wondering if we could have a FAQ on this topic
Some of my questions would be

-What are the things to taken care of during an Upgradation?
-Do's and Donts during an Upgradation(Installation)
-Cautions and Precautions during Migration
-Better methods of Export and Import
-Can multiple Versions be installed on the same system
-Is NLS enabling really required
-Options not to select which would later become a bottleneck

Datastage experts share their experiences and views on this!!!!
(You may have many more such questions to be added to this

)

kduke · Post by **kduke** » Fri Jun 24, 2005 4:52 pm

1.
a. Backup at OS level
b. Export all jobs in all projects.
c. Backup all critical DataStage files
i. dsenv
ii. uvconfig
iii. uvodbc
iv. odbc.ini
2. Do a root install
Don't do a dsadm install.
3. If you have PX or other IBM products then back them up too.
PX requires a database so have the Oracle client or whatever client
you need.
4. Download the DataStageBackup.bat from ADN.
Schedule it every night.
5. Yes mutiple versions can be installed but why do want to.
6. NLS is only needed if you have multiple languages in your
source data. You may have problems if you try to switch it
from NLS to non-NLS.
7. Upgrading with all plugins is slower but less trouble later.

This is my experience. I would always upgrade over a clean install.
Upgrades change your jobs as they go. If an upgrade fails in the middle
then it might not upgrade the jobs so try to make it go smooth.

kduke · Post by **kduke** » Fri Jun 24, 2005 4:57 pm

Here is a shell script to backup the critical UNIX files. I know this poster is on Windows but should still help.

Code: Select all

#!/bin/ksh
# run as dsadm
# this runs on Sun Solaris
# may not work on other UNIX
##
# may need to delete sudo lines
##
set -x
cd 
mkdir $HOME/SavedFiles
cp .profile SavedFiles
DSHOME=`cat /.dshome`
cd $DSHOME
sudo cp .[a-z]* $HOME/SavedFiles
cp dsenv* $HOME/SavedFiles
cp uvodbc* $HOME/SavedFiles
cp uvconfig* $HOME/SavedFiles
cd ../Projects
find . -name uvodbc.config -print | cpio -padmuv $HOME/SavedFiles
find . -name .developer.adm -print | cpio -padmuv $HOME/SavedFiles
find . -name .operator.adm -print | cpio -padmuv $HOME/SavedFiles
find . -name .prodmgr.adm -print | cpio -padmuv $HOME/SavedFiles
cd $ORACLE_HOME/network/admin
cp tnsname* $HOME/SavedFiles

# backup crontabs
mkdir $HOME/SavedFiles/crontabs
cd /var/spool/cron/crontabs 
sudo find . -group dstage -print | sudo cpio -padmuv $HOME/SavedFiles/crontabs
sudo find . -group dstage2 -print | sudo cpio -padmuv $HOME/SavedFiles/crontabs
sudo find . -group dstage3 -print | sudo cpio -padmuv $HOME/SavedFiles/crontabs

# exit
sudo cp /etc/passwd $HOME/SavedFiles
sudo cp /etc/group $HOME/SavedFiles
sudo cp /etc/shadow $HOME/SavedFiles

sudo chown -R dsadm $HOME/SavedFiles
sudo chmod -R u+rwx $HOME/SavedFiles

narasimha · Post by **narasimha** » Fri Jun 24, 2005 5:19 pm

That was informative!!
Is there any information or file which has to be manually copied from the old version to the new version.
One thing I heard was is that the Environment variables are not included during an import and the parameter file has to be manually copied to the new install, is that right?
Are there any such things to be taken care of !!!
Has anybody come across any post migration problems...

kduke · Post by **kduke** » Fri Jun 24, 2005 5:42 pm

Most of the files I backed up in this shell script are overwritten. So I copy them to the home directory of dsadm. It is easy to restore them back. Always when changing dsenv or one of these files then back it up in the directory with date stamp on the end.

cp dsenv dsenv.20050624

The parameters are also overwritten but we did not use them when I wrote this script. If you or someone adds them then please post the changes. Also if you change this to a batch file then post it as well. ODBC is handled differently in Windows and some of these do not need to be backed up on Windows. This is close enough to help you write a DOS version. This also does not backup the PX config files or any QualityStage config files.

ogmios · Post by **ogmios** » Sat Jun 25, 2005 7:52 am

NLS is also required in V7.5 and higher, or you get a warning message in every job.

Ogmios

chulett · Post by **chulett** » Sat Jun 25, 2005 8:01 am

What? Don't think so. I've got 7.5 installed on a test machine without NLS and have no such issue.

What makes you say that? Under what circumstances?

ogmios · Post by **ogmios** » Sat Jun 25, 2005 8:15 am

I got it a few months ago when I cold installed DataStage V7.5.1 on a Solaris F15K. Every job you started got a warning about NLS and hash files. Will check my old mails on monday for the actual error message.

I checked with Ascential support and I got an answer back saying that NLS should be switched on, and that the ability to switch off NLS is obsolete. And that the default characters set with NLS switched on is the same as for NLS switched off.

I reinstalled DataStage with NLS on that time.

Ogmios

chulett · Post by **chulett** » Sat Jun 25, 2005 8:24 am

Interesting. Maybe it's a 7.5.1 thing...

Anyone else have that experience?

ArndW · Post by **ArndW** » Sat Jun 25, 2005 8:25 am

Ogmios,

that statement from support scares the bejeesus out of me. Not only do the Americans generally not understand what NLS actually is but they have very little understanding of foreign characters and even what such things as collation and locales are. (Ask most people, including myself, where the French C-cedille is supposed to be collated and you would get at best a guess).

The internal handling of multibyte enabled strings is vastly different and almost an order of magnitude slower - at a assembly level the actions for MYstring[1000,2] are simple, offset 100 bytes from beginning of string location and remove the next 2 bytes. The same action when NLS is enabled, particularly with a non-stateless mapping, has to start at offset 0, read each character, detect if it is a starting sequence or a mid-sequence or a terminator (the method used depends upon what character set is in use), and repeat this until the 1000th character is reached. Big difference... Thus I would take any statement about NLS having to be used with about a bucketful of salt.

chulett · Post by **chulett** » Sat Jun 25, 2005 8:41 am

ArndW wrote:Not only do the Americans generally not understand what NLS actually is but they have very little understanding of foreign characters and even what such things as collation and locales are.

[raises hand] Guilty as charged, Your Honor... not having had a need to use it as of yet. Mon ami, I probably wouldn't know a C-cedille if it bit me on the butt.

Not to derail this thread any more than it already has been, but perhaps the subject of NLS and its use in / effect on DataStage might be the subject of a new FAQ? Or perhaps I should just polish my Google-Fu and start reading.

ArndW also wrote:The internal handling of multibyte enabled strings... <snip> ...and repeat this until the 1000th character is reached. Big difference... Thus I would take any statement about NLS having to be used with about a bucketful of salt.

Two bucketfuls.

kduke · Post by **kduke** » Sat Jun 25, 2005 9:39 am

We had 7.5.1 and NLS was not enabled. So I do not think this is true.

ogmios · Post by **ogmios** » Sat Jun 25, 2005 10:26 am

I remember the problem again... I activated NLS during installation but disabled it afterwards in uvconfig until we would need it for one or the other project. Then you get a warning in every job.

Ogmios

chulett · Post by **chulett** » Sat Jun 25, 2005 10:38 am

Ah. Something for the OP to make note of.