Transcript MIT - Unicode and OSDB Migraiton ASUG Oct 1, 2010

MIT Unicode
Conversion and OS/DB
Migration
]
[ JUERGEN LINDNER
SAP POINT OF CONTACT
MEMBER SINCE: 1998
/ Tikkana Akurati
[10.1.10
UnicodeSAP.com
[ LINDA WILSON
ASUG INSTALLATION MEMBER
MEMBER SINCE: 1999
[ ATUL PATANKAR
ASUG INSTALLATION MEMBER
MEMBER SINCE: 2000
[
OS/DB Migration and Unicode
Conversion of MIT SAP HR system
Tikkana Akurati
SAP Upgrades, Unicode Conversions and OS/DB Migration Specialist
UnicodeSAP.com
[email protected]
C: 847-804-2970
Real Experience. Real Advantage.
2
[ Project Goal




Retire current SUN hardware
Migrate to faster and cost-effective Hardware running Linux
Unicode convert while migrating to new hardware.
Complete the Migration and Unicode Conversion over a long
weekend
 5/28/10 – 5/31/10 (Memorial Day Weekend)
Real Experience. Real Advantage.
3
[ Presentation Agenda











System Information
Database Stats
Conversion And Migration Stats
Tools and Techniques used
Distribution Monitor Setup
Progress Monitors
Target Unicode System Tablespace Summary after Conversion and
Migration
Checks (Basis)
Check for Missing Objects and Indexes
Target Unicode System Tablespace Summary and Missing Objects in
SAP check
Issues and Go Live feedback
Real Experience. Real Advantage.
4
[ System Information
 Hardware Information Involved in Unicode Conversion and OS/DB
Migration
 Non-Unicode DB
 SunFire V890 with
 Memory size: 28 GB
 8 CPU x 1.8GHz
 Unicode DB
 Dell R900 with
 Memory size: 64GB
 16 CPU x 2.9GHz
 One Additional Server used for Conversion
 Dell R900 with
 Memory size: 64GB
 16 CPU x 2.9GHz
 Operating System Software
 Solaris 9
 Red Hat Enterprise Linux Server release 5.2 (Tikanga)
 Old SAP Codepage: 1100
 New SAP Codepage: 4103
Real Experience. Real Advantage.
5
[ Database Stats
Source Database
Target Database
Total Used Size
723GB
632GB (12.5% reduction)
# of tablespaces
36
7
DB - Codepage
US7ASCII
UTF8
DB version
10.2.0.4.0
10.2.0.4.0
Largest 5 tables
PCL2
48GB
RFBLG
30GB
PCL4
28GB
BSIS
23GB
COEP
22GB
Real Experience. Real Advantage.
6
[ Conversion And Migration Stats







Total number of export packages
Number of splitted tables
Number of additional import packages
Packages with special handling
Total Conversion Time
Target Database shrunk by
# of issues encountered
Real Experience. Real Advantage.
:
:
:
:
:
:
:
1041
42
84
SAPVIEW
8hrs 50mns
12.5%
1 (PCL2-41)
7
[ Tools and Techniques Used
 All the tools used are standard and provided by SAP
(except a few custom scripts to monitor Oracle and build target DB tablespaces.)
Distribution Monitor
Table Splitter and WHR Splitter
Package Splitter
OrderBy
Oracle parameter optimization for export/import: Note
936441
 Sapinst
 General Unix monitoring tools
 Custom Oracle SQL scripts to build target DB faster and to
monitor Oracle.





Real Experience. Real Advantage.
8
[ Distribution Monitor Setup
 # of Servers Performing Conversion: 2
 # of R3load processes on each: 10 to 20 (Adjusted dynamically
based on CPU and Disk usage)
 Source DB on Solaris acted as DB server only.
 Target DB on Linux acted as DB server and also perform
conversion.
 One application server (Linux) configured to perform
conversion only.
 # of tables split: 42
 # of tables treated as individual packages: 620
Real Experience. Real Advantage.
9
[ Distribution Monitor Design
AppServer1
Export
Source Server
Non-Unicode
DB Server
Linux
Import
/EXP
Common NFS
export dump
Target server
Unicode DB
Server
Solaris
Linux
AppServer2 (Same as Target)
Export
Real Experience. Real Advantage.
Linux
Import
10
[ Progress Monitors

last refresh: 2010-05-29 11:32:03
|-----------------------------|
| SYSTEM | SID | DBTYPE |
|-----------------------------|
| SOURCE | PS1 |
ORA
|
| TARGET | PS1 |
ORA
|
|-----------------------------|
|----------------------------------------------------------------------|
|
ACTION
| TOTAL | INITIAL | RUNNING | COMPLETED | FAILED |
|----------------------------------------------------------------------|
| JUGLANS EXPORT
|
420 |
|
|
ALL |
|
| JUGLANS IMPORT
|
454 |
|
|
ALL |
|
|----------------------------------------------------------------------|
| CARYA EXPORT
|
621 |
|
|
ALL |
|
| CARYA IMPORT
|
671 |
|
|
ALL |
|
|----------------------------------------------------------------------|
SAPVIEW import finished on host JUGLANS


Unix commands : top, iostat
Oracle
: V$SESSION_LONGOPS, Used database size (dba_tablespaces)
Real Experience. Real Advantage.
11
[
Target Unicode System Tablespace
Summary after Conversion and Migration
Real Experience. Real Advantage.
12
[ Checks (Basis)
 Check row counts of some critical tables and compare
between Non-Unicode source and Unicode target systems
 Check for missing Indexes and data inconsistencies (Shown as
example on the next slide)
 Check system Status for Unicode and SP Levels
 Check all the export and import logs for any errors
 Check for completion of all packages in distribution monitor
status and logs
 Check and compare DB sizes before and after conversion.
Real Experience. Real Advantage.
13
[ Check for Missing Objects and Indexes
Real Experience. Real Advantage.
14
[ Issues & Golive feedback
 Issue during Conversion with sqlnet timeout – Solved by reexporting again – PCL2-41 table split with 12 records.
 One variant with “not current version” status – Solved by
running RSVARDOC_610
 “we ran a full monthly payroll simulation run in the new
production environment. We ran a similar run on 3/15/2010 in
production; it took 6749 seconds to complete – the run we
executed today only took 2975 seconds, a 56% processing
speed improvement.”
Real Experience. Real Advantage.
15
[ Issues Post Go-Live
1.
Some outbound extracts failed because could not TRANSFER newly input Unicode
characters (long dash, fancy quotes, etc) in SAP into non-Unicode outbound file formats.
a.
This was one of our biggest issues. Although we tested all our critical outbound feeds, we did
not do thorough enough end-to-end tests of inputting unusual (non ASCII) Unicode character
into SAP and then testing the downstream outbound feeds with that specific data.
b. For our near term fix, we’ve chosen to use the OPEN DATASET … IGNORE CONVERSION
ERRORS extension replacing either with a “#” or “ “ (space) depending on the recipient system.
c. 100+ programs have been updated, about 70 more to go. Only about 10 actually short
dumped because of Conversion Codepage error. The rest are proactively being changed.
2.
Vendor firewalls – IP address of new hardware needs to be communicated to vendors
3.
Hardcoding of the old batch server name – since we did not have file sharing, a few
programs and variants had hardcoded “beech” in them
4.
HR log data could not be read (importing from cluster tables). Had to delete pre-6/1
data out of the HR log tables so new log data could be read.
5.
When we converted to RedHat Linux, we also converted to GPG encryption (from
PGP). While most errors were caught in testing, we did have 1 post go live issue. GPG
and Linux is far more strict (less forgiving) than PGP on Solaris.
Real Experience. Real Advantage.
16
[ Q&A
Real Experience. Real Advantage.
17
[
]
 Thank you.
Tikkana Akurati
SAP Upgrades, Unicode Conversions and OS/DB Migration Specialist
UnicodeSAP.com
[email protected]
C: 847-804-2970
Real Experience. Real Advantage.
18