Category Archives: Migration
Today signals the end and the beginning for a year-long project. My project for SQL Server Domination. Our shop is primarily an Oracle shop with dot net developers. Although I do have almost 300 databases across all levels, the bulk of mission critical applications are written for Oracle.
A little over a year ago we began a plan to migrate our financial applications to SQL Server in order to retire the oldest Oracle servers on site and reduce our massive licensing footprint. This was all part of my plan for SQL Server World Domination (said in a maniacal voice, imagine that). We knew that if this mission critical migration was a success that we could leverage our product against the Oracle servers in-house and begin the destruction of their infrastructure in favor of my glorious SQL Servers! (Hear the trumpets heralding the new king?)
This project, if successful, will be the Oracle death knell. Over the last year I have prepared diligently with one goal in site: this migration weekend. The weekend of my birthday seems apropos to give me the present that I so desire, SQL Server Domination. You may have noticed that the number of blog posts recently have been limited especially after blogging every day last year. For the last few months, I have had one purpose and one purpose only: ensuring the success of this migration!
Let the trumpets herald the coming of a new dawn. Long live SQL Server. I will blog again next week on the other side of this project to declare that the Queen is dead and a new King has ascended the throne! Until then queue the Imperial March, hello Starbucks and massive four day working weekend with little sleep migration project here I come!
Yesterday I was tasked with migrating our Citrix EdgeSight database off of a 2005 SQL Server and migrating it to a new 2008R2 Server. No problem, I have done that a million times before. But wait a minute, that database has SSRS reports installed on that server and not our main reporting server. If you follow my blog, you will probably see that I have not had much experience with SSRS, my reporting experience has been with Crystal and other minor packages.
I started researching how to migrate these reports and was not pleased to see Citrix’s response. They recommended migrating the ReportServer and ReportServerTempDB databases to the new server and then reinstalling SSRS on the original server. So what happens then if you have other jobs on the new and old servers? This is not a good recommendation. After an exhaustive search the only option I saw within the tool was to download reports one at a time and upload them into the new server. This was also not an option with a fair amount of reports to migrate.
Enter Reporting Services Scripter from SQLdbatips.com. Run the program on the original server and select your Report folder and scripts are generated and placed into a folder of your choosing. These scripts can then be copied over to the new server and executed by command shell. In this particular instance we had to reopen the data source and enter new credentials. Problem solved, reports migrated successfully.
I would love to upgrade my servers to SQL Server 2012. We’ve met with Microsoft and we discussed the licensing ramifications and I got really excited about upgrading after getting the OK internally. That was a couple of months ago. The one piece holding us up? Vendors!
We do have some SQL databases developed in-house, but being predominantly an Oracle shop, most of our SQL Server databases are vendor provided. We have set forth the initiative to do all new development solely in SQL Server for cost savings. Additionally, we have begun moving some of the vendor provided databases to SQL Server. However, those are also at the mercy of vendors. My latest project only supports SQL Server 2008, not even R2. This is very frustrating from an enterprise DBA standpoint. I want to move forward! Enjoy!
After using the SQL Server Migration Assistant for Oracle on a current project to migrate an Oracle database to SQL Server, I have learned a couple of little tricks that I would like to pass on to you. These are mainly time savers so hopefully they will make you more productive in your project.
The first problem I noticed is that there was no easy way to copy projects nor was there a way to save a project file as another name, which would have allowed me to copy the project. Why would I want to copy a project? There are many settings such as column type mappings that will stay the same for my project and are very time-consuming to enter, if you are modifying the defaults. In my migration, I want to do many iterations in my development and acceptance environments with different SQL Server database names such as Round1, Round2, or Round3 and then use Red Gate Compare and Data Compare to verify that all of the schema and data from the migration are consistent in my migration. This gives me a warm and fuzzy feeling before I turn over the data to the testers.
However with each iteration, I must create a new project manually. Why? Because SSMA does not refresh to recognize new databases in the existing projects. That is very annoying. It only shows what is there at the time the project is created. Even if you reconnect to the SQL Server, it does not refresh.
If you know ahead of time that you will do many iterations, you could however create several databases before starting with the SSMA and then copy the project folder from the first project that you create manually. You must then rename the folder to the new project name and the corresponding project file (with an o2ssproj file extension) contained within the project folder. This would allow you to simply switch to the new database. You must also keep in mind that you need to double-click on the Oracle schema and make sure the target schema is selected as your new database. This will not happen automagically as you choose a new database name in the SQL Server metadata explorer.
Another option is to create new projects as you create the databases and then copy in the object-container.mappings and/or project-container.mappings files from the first project folder into the new project folder. This will allow you copy those settings into the new project and save you some time. I hope you find this useful. Enjoy.
The Oracle Conversion, day 14, continues and it is not as exciting as I envisioned. Oh, don’t get me wrong, the end result, Global Domination for SQL Server, will be the ends that justify the means. But until then, migrate data, check data, clean data, remigrate date, check data, clean data, time migration, repair migration process, check data, clean data, remigrate, ad nauseam. Rinse, lather, repeat.
Oh, this is not my first rodeo, I’ve converted many projects before but never have I converted Oracle before. Granted, the SQL Server Migration Assistant for Oracle has made life much easier but at the same time it has made life much more difficult.
It is a quirky tool that is very temperamental. We had to restart it before we did a successive migration or it would fail we are guessing because it possibly caches information from the previous migration process and the MSDN blog wasn’t much help. We also had to hit apply on every setting change or it would revert some of them back to the original value.
In addition, we would have to convert schema sometimes four or five times before it would synchronize them properly to SQL Server, saying that the SQL execution failed. That answer gives me a huge amount of troubleshooting possibilities. When it works, it was a great time saver but had we known the quirkiness we probably could have scripted it all out and saved a lot of headaches.
Doesn’t sound like a big deal?
Well when you have some big tables and it takes eight hours to convert the schema and you try synchronizing and it fails and you have to convert the schema again, you watch the progress bar for many hours. Oh yeah, I left out the fact that the vendor is responsible for the migration contractually, so I guess ultimately they should have scripted it out. Enjoy!
For the last two weeks, I have been working with a vendor to migrate our financials from Oracle to SQL Server using SQL Server Migration Assistant. This has been a time-consuming process, which requires an extreme attention to detail as precious data hangs in the balance.
We have found numerous pieces of ‘bad data’ where either Oracle or the application allowed date such as 03/02/0007. These dates are allowed to migrate to SQL Server when using a datetime2 with precision 0, however the vendor requested to use datetime originally which ranges from 1/1/1753 to 12/31/9999, thus causing migration errors.
Here is the dilemma, do you modify data for financials that have closed to have clean data or leave data as is? The DBA in me says to clean the data and make everything pristine before you move it into my SQL Servers, however there are implications with financials and the department made the decision to import as is to ensure data was identical and unmodified. That I can certify, well the vendor will need to certify that as they have chosen the column type conversion fields that work with their application’s SQL Server version.
This conversion is a critical piece in my plan for global domination, well that is my plan to destroy our dependence on Oracle and move solely to SQL Server especially since we are a dot NET development shop. When this project is successful and users see better performance then we will be able to migrate other databases as well and we will see better performance! Enjoy.
You’ve migrated your database to a new box, with no problems. You created the login that goes with this database, and all is right in the world. However, when your user goes to connect to it they get the dreaded error 18456, login failed. But wait, I created the login, right?
Not so fast there, accidental DBA. You are going to need the help of a nifty stored procedure called sp_help_revlogin. This stored procedure, provided by Microsoft in the link below, will generate a T-SQL script that will copy the logins, their passwords and their corresponding SIDs (or security identification number, which is usually the culprit here). This is also a great utility when the database has numerous users with various security levels and passwords. Really now, who wants to write all of that down and recreate it? Even with screenshots, it can turn into a large, fat-fingering and time consuming event. Mind-numbing, I would say. Why do something manually when you can do it automagically? Just think of all those happy logins that will not be orphaned by the dark side?
Here is also a code snippet to check for orphaned logins, for your enjoyment. Enjoy!
USE <database_name>; GO; sp_change_users_login @Action='Report'; GO;