Question Dump & Load Question

nix1016

New Member
We're upgrading all of our clients from version 10.2B to 11.6 and one of the processes involved is to convert the databases from ISO8859-1 to UTF-8. The process I'm currently doing is:

1. Convert version 10 DB to UTF-8
2. Binary dump version 10 DB
3. Load .bd to empty version 11 UTF-8 DB

The thing is that after the conversion to UTF-8, it requires an index rebuild which takes a while on a large database. After loading the .db files, it again requires an index rebuild.

My questions is if I skip the index rebuild after converting to UTF-8, can I still do a binary dump without any issues and simply rebuild the indexes in the version 11 DB after the load? It will be a great time-saver if this is possible without encountering any issues.

Also, is there a way to find out the database code page programmatically (I can't seem to find this in the knowledge-base)? I'm trying to streamline the process in a script and I need to prevent the dump from happening if the database hasn't been converted to UTF-8.

Thanks in advance!
 

RealHeavyDude

Well-Known Member
I am not aware that an index rebuild is necessary when you change the database code page with proutil convchar convert.
Progress KB - How to convert an existing database to Unicode UTF-8?
See method without an index rebuid.

To find out the database codepage programmatically you can run an ABL procedure against the database that writes the oupput of the dbcodepage function into a file that gets read in your script.

Heavy Regards, RealHeavyDude.
 

cj_brandt

Active Member
The Progress KB states 1 idxbuild if you use convchar and 2 if you don't.

In 10.2B08 there are many performance improvements made to the idxbuild utility. If your index rebuild was taking several hours, you could see a 50% reduction or more in time. The improvements first appeared in 10.2B0606, but there were a few bug fixes added to SP08.

There is an enhancement request to find the code page via VST.
 

nix1016

New Member
Thank you for that, it seems I need to use idxbuild regardless of which method I choose to convert to UTF-8? Unfortunately most of our old clients (the ones with the big databases) are currently on 10.1C04.

Yes I had hoped that there was a VST that could tell me the code-page but dbcodepage will work for me. Thanks!
 

nix1016

New Member
I have another question in regards to this, after I convert the version 10 DB to UTF-8, I'm loading the ICU-UCA.df collation file to replace the basic collation. I can't seem to get rid of the warning that pops up after loading the collation file telling me to rebuild the index. Is there a way to suppress this warning? I have tried using output around the run prodict/load_df.p call but it still pops up with the prompt.

If not possible, then will it cause any issues if I don't apply the collation but dump the data in basic collation but load it into a version 11 DB with ICU-UCA collation applied?
 
Top