Using SQL Source Control with Fortress or Vault – Part 2
In Part 1, I started talking about using Red-Gate’s newest version of SQL Source Control and how I really like it as a viable method to source control your database development. It looks like this is going to turn into a little series where I will explain how we have done things in the past, and how life is different with SQL Source Control. I will also explain some of my philosophy and methodology around deployment with these tools. But for now, let’s talk about some of the good and the bad of the tool itself.
More Kudos and Features
I mentioned previously how impressed I was with the responsiveness of Red-Gate’s team. I have been having an ongoing email conversation with Gyorgy Pocsi, and as I have run into problems or requested things behave a little differently, it has not been more than a day or two before a new Build is ready for me to download and test. Quite impressive!
I’m sure much of the requests I put in were already in the plans, so I can’t really take credit for them, but throughout this conversation, Red-Gate has implemented several features that were not in the first Early Access version. Those include:
- Honoring the Fortress configuration option to require Work Item (Bug) IDs on check-ins.
- Adding the check-in comment text as a comment to the Work Item.
- Adding the list of checked-in files, along with the Fortress links for automatic History and DIFF view
- Updating the status of a Work Item on check-in (e.g. setting the item to Complete or, in our case “Dev-Complete”)
- Support for the Fortress 2.0 API, and not just the Vault Pro 5.1 API. (See later notes regarding support for Fortress 2.0).
These were all features that I felt we really needed to have in-place before I could honestly consider converting my team to using SQL Source Control on a regular basis. Now that I have those, my only excuse is not wanting to switch boats on the team mid-stream. So when we wrap up our current release in a few weeks, we will make the jump. In the meantime, I will continue to bang on it to make sure it is stable. It passed one test for stability when I did a test load of one of our larger database schemas into Fortress with SQL Source Control. That database has about 150 tables, 200 User-Defined Functions and nearly 900 Stored Procedures. The initial load to source control went smoothly and took just a brief amount of time.
Warnings
Remember that this IS still in pre-release stage and while I have not had any problems after that first hiccup I wrote about last time, you still need to treat it with a healthy respect. As I understand it, the RTM is targeted for February. There are a couple more features that I hope make it into the final release version, but if not, they’ll probably be coming soon thereafter. Those are:
- A Browse feature to let me lookup the Work Item ID instead of having to remember it or look back in my Item details. This is just a matter of convenience. I normally have my Work Item list open anyway, so I can easily look it up, but hey, why not make it even easier.
- A multi-line comment area. The current space for writing check-in comments is a single-line text box. I would like to have a multi-line space as I sometimes write lengthy commentary. But I recognize that it is a struggle to get most developers to put in more than the word “fixed” as their comment, so this meets the need of the majority as-is, and it’s not a show-stopper for us.
- Merge. SQL Source Control currently does not have a Merge feature. If two or more people make changes to the same database object, you will get a warning of the conflict and have to choose which one wins (and then manually edit to include the others’ changes). I think it unlikely you will run into actual conflicts in Stored Procedures and Functions, but you might with Views or Tables. This will be nice to have, but I’m not losing any sleep over it. And I have multiple tools at my disposal to do merges manually, so really not a show-stopper for us.
Automation has its limits. As cool as this automation is, it has its limits and there are some changes that you will be better off scripting yourself. For example, if you are refactoring table definitions, and want to change a column name, you can write that as a quick sp_rename command and preserve the data within that column. But because this tool is looking just at a before and after picture, it cannot tell that you just renamed a column. To the tool, it looks like you dropped one column and added another. This is not a knock against Red-Gate. All automated scripting tools have this issue, unless the are actively monitoring your every step to know exactly what you are doing. This means that when you go to Deploy your changes, SQL Compare will script the change as a column drop and add, or will attempt to rebuild the entire table. Unfortunately, neither of these approaches will preserve the existing data in that column the way an sp_rename will, and so you are better off scripting that change yourself. Thankfully, SQL Compare will produce warnings about the potential loss of data before it does the actual synchronization and give you a chance to intercept the script and do it yourself.
Also, please note that the current official word is that SQL Source Control supports Vault Professional 5.1 and later. Vault Professional is the new name for what was previously known as Fortress. (You can read about the name change on SourceGear’s site.) The last version of Fortress was 2.x, and the API for Fortress 2.x is different from the API for Vault Pro. At my company, we are currently running Fortress 2.0, with plans to upgrade to Vault Pro early next year. Gyorgy was able to come up with a work-around for me to be able to use SQL Source Control with Fortress 2.0, even though it is not officially supported. If you are using Fortress 2.0 and want to use SQL Source Control, be aware that this is not officially supported, but it is working for us, and you can probably get the work-around instructions from Red-Gate if you’re really, really nice to them.
Upcoming Topics
Some of the other topics I will likely cover in this series over the next few weeks are:
- How we used to do source control back in the old days (a few weeks ago) before SQL Source Control was available to Vault users
- What happens when you restore a database that is linked to source control
- Handling multiple development branches of source code
- Concurrent Development practices and handling Conflicts
- Deployment Tips and Best Practices
- A recap after using the tool for a while