Database collation importance - HIGHLY TECHNICAL HELP NEEDED

Sort:
You are not authorized to post a reply.
Author
Messages
John Costa
Veteran Member
Posts: 154
Veteran Member

    First, I apologize if this thread was started in the wring forum, but I didn't really see where else it would fit.

    I know this is likely to be a highly technical issue, but I sure hope someone can provide a detailed explanation for me.
     
    How important is the specified collation for a database, specifically a Lawson database?
     
    When our 8.0.x environment was installed for us back in 2004 (by Lawson, I might add), they also installed SQL-Server 2000 as part of that install. However, the installer did not specify a default collation to be used for SQL-Server. As a result, SQL-Server was installed with a default collation of SQL_Latin1_General_CP1_CI_AS. This also resulted in our LOGAN and application product line database (named PROD1) being set up to use this same default collation.
     
    Now fast forward to 2009. I recently completed an LSF9 installation on my new application and database servers. However, I did not install SQL-Server on my database servers, my DBAs did. Unfortunately they were not aware that, according to the LSF9 installation documents, we should be using a default collation of Latin1_General_BIN. So, SQL-Server was again installed using the default collation of SQL_Latin1_General_CP1_C1_AS. When I did my LSF9 install, I specified a collation of Latin1_General_BIN for the GEN and LOGAN databases as dictated by my documentation. The DBAs then simply restored a backup of my PROD1 database onto the new database server. 
     
    So, here is where I’m at. I have SQL-Server, along with my PROD1 application database, using a collation of SQL_Latin1_General_CP1_C1_AS while only my LOGAN and GEN databases are using a collation of Latin1_General_BIN. I’m now running into issues with interfaces that migrate data into and out of my various databases because of differences in collation.   Granted, I’m pretty good with SQL and can code around these problems, but it sure would be a lot simpler if all my databases were using the same collation from the get-go.
     
    So back to my question: How important is the specified collation for a database, specifically a Lawson database? Can I rebuild by LOGAN and GEN databases so that they use the same collation the rest of the server is using?  Changing the collation on my PROD1 database is not an option simply due to the size of the database and the length of time that would be needed.
     
    I know how to change a database collation; that is not the concern I have. What I really need to know is why Lawson has specified a collation of Latin1_General_BIN? Is there a specific reason behind it or can I use another collation on my GEN and LOGAN databases with no adverse effects?
    Thanks in advance for any help you can provide.
    _________________ John - Wichita, KS
    John Henley
    Senior Member
    Posts: 3349
    Senior Member
      John, you definitely DO NOT want to change the other database to SQL_Latin1_General_CP1_C1_AS.
      The reason why Laton1_General_BIN is used is because, amongst other things, it is case-sensitive. In the case of SQL_Latin1_General_CP1_C1_AS, the "CI" means it's case insensitive. Where this causes problems in particular (I've seen this before at some clients and have figure it out, and rectify it) is that you start getting all kinds of strange GEN and LOGAN errors for fields that are indexed and have duplicate values if looked at without case (i.e. Gbi is GBI is GBi, etc.) That is a real-life example, when a client moved to LSF9, there GEN database was set up incorrectly, and they had errors galore--for example srgen would no longer run because "duplicate" key numbers were in the database. Changing the collation on the database is not an easy task (btw, "changing" it on the database doesn't actually "fix" existing data).
      Thanks for using the LawsonGuru.com forums!
      John
      John Costa
      Veteran Member
      Posts: 154
      Veteran Member
        Thanks for your help, John. I figured as much but I was hoping there was a way around it. Looks like we have quite a task ahead of us in changing the collation for our application database. The ACTRANS table alone took 8 hours to migrate from one database to another.
        _________________ John - Wichita, KS
        John Henley
        Senior Member
        Posts: 3349
        Senior Member
          Curious--roughly how many rows in ACTRANS? Also, what apps version are you running? If you're still on 8.0.x or 8.1, you can combine this with apps upgrade, which is what I did with one client who had a huge database and needed to fix collation.
          Thanks for using the LawsonGuru.com forums!
          John
          John Costa
          Veteran Member
          Posts: 154
          Veteran Member

            We are on apps 8.1, MSP6.  As much as I would like to upgrade the Apps, the boss prefers we do that upgrade later this year (along with going from LAUA to Lawson Security).

            Here's a breakdown of our tables, top 10 largest: (in row count)

            ACGLACCT - 109,549,863
            ACMXVAL - 74,644,904
            GLTRANS - 52,916,396
            ACTRANS - 52,243,844
            GTMXVALUE - 37,445,824
            ACOVERRRIDE - 33,669,456
            PRDISTRIB - 17,065,374
            PRTIME - 11,496,293
            ACBRDNTRAN - 10,789,907
            ACMASTER - 10,558,615

            Depressing, isn't it?  I should have listened to my mother and become a truck driver....

            _________________ John - Wichita, KS
            John Henley
            Senior Member
            Posts: 3349
            Senior Member
              Hmm. Looks like you're not utilizing new AC structure features in 8.1 :)
              Thanks for using the LawsonGuru.com forums!
              John
              John Henley
              Senior Member
              Posts: 3349
              Senior Member
                You can definitely do it as part of the apps upgrade process. You should probably change your database default collation in the meantime so that new databases (assuming this SQL server instance only supports Lawson databases) are created with Latin1_General_BIN. Since most Lawson tables/forms uppercase the key values in a table (an exception is requesters), there is only a slight risk of key collisions, and if you haven't run into it since your install in 2004, you're probably safe to leave it as is until you do the apps upgrade. It can be done, however, but requires doing it table-by-table, column-by-column, using "ALTER TABLE XX ALTER COLUMN YY COLLATION " statements.
                Thanks for using the LawsonGuru.com forums!
                John
                John Costa
                Veteran Member
                Posts: 154
                Veteran Member

                  John - Since our LOGAN and GEN databases are using the correct collation (Latin1_General_Bin) and our application product line (PROD1) is not, what potential problems might we run into going forward?  Is there a potential for any kind of data corruption because of the differences in collation?  I ask because I don't know what table joins go on in the background between LOGAN, GEN, and PROD1.

                  As I mentioned earlier, we plan to upgrade to apps version 9 later this year.  Will our application product line need to be using the correct collation before we embark on that upgrade project or can we do the collation change during the application upgrade?

                  On the surface of this whole mess, you might think just changing the collation now would be the right thing to do.  I agree that we should be doing it now and get our SQL-Server 2005 platform "speaking the right language."  It's the application outage this would require that's preventing us from moving forward.

                  I'd appreciate any further suggestions you might have to offer.

                  _________________ John - Wichita, KS
                  John Henley
                  Senior Member
                  Posts: 3349
                  Senior Member
                    There are no direct table joins between the apps product line and LOGAN and GEN. Where you may have difficulty is with forms/tables that allow mixed cases in index fields. For example, I think RQ04/requester allows you to define a requester in upper or lower case, so you could have JIM and Jim and jim. Since you have your apps product line defined with case-insensitive, this would be a "key collision".

                    Since you have been running since 2004 with the incorrect collation, chances are good that you will not run into any issues between now and your upgrade. To analyze any potential issues would require creating a SQL script to analyze all table columns that are included in indexes, and looking for duplicates in their upper-cased values, e.g. something like (off the top of my head):

                    SELECT UPPER(REQUESTER), COUNT(UPPER(REQUESTER)) FROM REQUESTER GROUP BY REQUESTER HAVING COUNT(UPPER(REQUESTER))>1

                    When you do the apps upgrade, assuming you follow the standard procedure of copying PROD to PROD8, you would define both PROD8 and PROD9 as case-sensitive, then use sqldbcopy to copy data from PROD to PROD8 (don't just copy the physical database and reattach). This will put all of the data in PROD8 into the correct collation.
                    Thanks for using the LawsonGuru.com forums!
                    John
                    You are not authorized to post a reply.