Hi All,
Good Day.
Currently we have database in 10.1C progress version on HP-UX , we are planning to move to progress 11.5 on
Linux OS.
Since it includes Progress and hardware migration I created database with below steps.
1.prostrct create <db-name> db-name.st (FTP ed .st file from HP-UX to Linux machine)
2.procopy /usr/dlc115/empty8 <db-name>
3.Took all ..df and .d files from HP-UX and loaded into linux server .
So new database are created now. But when I check for DB size with Promon <db-name> and chosing Option 5.
there is difference in the database size.
Database Name HP-UX DB SIZE LINUX-DB SIZE
symglbl 1461MB 781MB
SSOEM 7052 MB 5130 MB
So after seeing the this difference , I am worried whether my new database on linux machine are accurate or not.
Please provide your thought on this.
Please let me know how can I do some test to ensure my new database on linux are same as that of database on HP-UX.
A dump and load procedure should always include a verification of record counts from tabanalys or dbanalys reports from the source and target databases to ensure all tables are accounted for and have equal record counts. Check that first.
Also, and this is less about the size difference and more about best practice, ensure that any other required data in the source databases (_User records, _sec-* records, SQL permissions, current sequence values) has been dumped and loaded as well.
Depending on the state of the source databases, it is possible that just the act of dumping and loading and rebuilding the indexes reduced the overall size of the databases by the amount you have measured.
Hi Rob,
I ran dbanalys on both hp-ux and linux database. I compared the record count column and it looks both have the same number of records .
I have loaded other stuffs like
slq-views
user table contents
security permisssion
sequence current values
Database Identification Properties
Database Options into linux database .
Then shall I asssume that database in hp-ux and database in linux are identical. Please provide your thought
Did you compare index size (BTW, did you binary D&L + idxbuild)?
Did you compare free space?
Did you compare % of the RM block space is used?
Hi George,
I used data dictionary to perform Dump and load .
Can u please tell where i can check for free space and % RM block space used?
"RECORD BLOCK SUMMARY" segment ends by the messages like:
2046047 RM block(s) found in the database.
97.91% of the RM block space is used.
Dbanalys ends by "DATABASE BLOCK ANALYSIS" segment with the messages like:
42188 free block(s) found in the database.
85 empty block(s) found in the database.
2579804 total blocks found in the database.
There are many reasons why the "size" shown in PROMON might differ. Two really big potential reasons would be changes in the rows per block and storage area type. Those might be deliberate changes or accidental changes. I suggest that you compare the .st file from HPUX to the .st file on Linux. Get current .st files using "prostrct list".
Do not compare the size using promon. Since you did dbanalys on both hp-ux and linux databases, compare those sizes. They should match a little better. And new database on linux should be smaller. All indexes were compacted ( rebuilt ). The unique indexes could shrink several times, I have seen 25:1 ratio index shrink after big table purge. It is actually very much recommended to do idxcompact after a purge. You could also have big RM chains in old database, that was taken care during D&L.
And finally compare the speed that dbanalys took on hp-ux vs linux. I bet it was much faster on new db. And thus database will work faster too.
> And finally compare the speed that dbanalys took on hp-ux vs linux. I bet it was much faster on new db. And thus database will work faster too.
A bit disagree. Indeed an application and dbanalys can run faster after D&L but in the different degree. And an application can work slower after D&L. Also it's important to differ the online and offline dbanalys. Their times may significanly differ one from another.