SQL Change Automation and GitLab CI/CD (a.k.a. Oh this is fun on Windows)

“You never know what you can do until you try, and very few try unless they have to.”
C.S. Lewis

Well I don’t have to, but many of the people I speak to on a daily basis are moving into GitLab, so it’s about time I tried it! You can find here testament to the mistakes I make as I try to set up a full end-to-end database change management process with SQL Change Automation and GitLab.

Will it all work perfectly? I don’t doubt that everything will fall over at some point, but let’s see how we get on all the same, and hopefully if you’re setting up this same pipeline, you’ll be able to avoid the errors and failings I inevitably cause! So here we go!

ready come on GIF

Let’s set up a GitLab Project (and rename the default branch)

Naturally, I didn’t have a GitLab account, so I had to set one up. I’m assuming that if you’re using it already or you’ve just started using it you’re taking advantage of the more business features but I’ve just stuck with the good ol’ free account for now! It was remarkably simple, sign up, email address, confirm and here we are:

Ok there is something very cool I like about setting up a new project, can you tell what it is?

You can completely set up a new blank project but they have templates, you can import projects OR, and I love this, you can setup a full CI/CD pipeline from another repo! Having done this before in Azure DevOps it was not easy, let me tell you. It really seems like Azure DevOps hates you for setting up CI/CD from an external repo, even though it has plenty of helpful ways of doing so!

So I initialized my repository with a README and updated it:

Don’t ever say I’m not descriptive enough!

The first thing I did was a renamed my default branch to ‘trunk’ by going to branches, creating the new branch and then in Settings > Repository changing it to the default and then swapping out the protected status with the outdated master:

Then finally delete the old default in Repository > Branches:

Excellent. Now it’s time to clone trunk onto my machine as we will need the local repository to put our change automation project in!

I created a folder called GitLab test and cloned the mostly empty repository into it:

Simple!

Create a new SQL Change Automation project and push it to trunk

In SSMS I opened up the most recent version of SQL Change Automation an created a new project called “DoggosAreCoolDB” using a copy of a Dev database I had lying around from a previous demonstration (BlogsDotRedgate):

Then I created my baseline as a migration script against the up-stream copy, BlogsDotRedgate_Integration, because who has access to Prod for this? Am I right? *cough & shifty eyes* not me!

I successfully generated my baseline and a change script (I added a column to a table, nice and simple) and then committed them to my local repo, and pushed! Forget branching, today isn’t about that, we’re just PUSHING TO TRUNK, WOO-HOO!

Setting up the CI/CD Pipeline

Now that we have our project and migrations in GitLab we can build out a pipeline! So first stop I went straight to CI/CD > Pipelines and was presented and I hit “Get Started”:

They immediately throw you into a Quick Start “Help” style guide which is immediately a little un-intuitive but surprisingly helpful if you read the whole thing. Effectively we need a YAML file called .gitlab-ci.yml that will store our pipeline as code telling it how and where to build, and we need a runner to actually fire up and execute these steps.

In my experience with some other CI/CD tools, it’s been advantageous to actually create the Runner / Agent first on the machines you’re going to be using, so as I just have my laptop to do this on, I will set one up on there! I found the full documentation for a Windows Runner here, and followed it just so I would have it available.

The GitLab Runner was up and running in my services but I’ll be darned if I can see them anywhere in GitLab…

confused britney spears GIF

Aha! So it turns out after a bit of digging that you need to register the runner specifically using the CI/CD section on the project settings, that was probably my bad for not reading the documentation thoroughly but my counter-argument… who actually does? So I issued the register command, applied tags and a description and chose my runner type, I chose shell because I need to be able to run PowerShell on the machine (I’ll need the SQL Change Automation PowerShell components available on the machine where the Build and Deployment are happening of course):

I’ve got the runner on the machine, I’m using an instance of SQL Server to build against, now I just need the YAML file (fortunately GitLab has full documentation for how to structure this as well!)

So I can build my project I’m going to need to know where the repo is cloned to during the process (i.e. to find the .sqlproj file) so by taking a look I managed to find a list of environment variables that can be used in the YAML file, just to be sure though, I created and committed the most basic YAML file that would just echo back the location of the cloned files:

stages:
  - DatabaseBuild

Build:
  stage: DatabaseBuild
  tags: 
   - sql
  script: 
   - echo $env:CI_PROJECT_DIR

After this let me know the environment variable worked correctly and the build pipeline was being fired up correctly on my private runner, I tried something a little more ambitious, building the .sqlproj file using the cmdlet reference from the SQL Change Automation documentation for help:

stages:
  - DatabaseBuild

Build:
  stage: DatabaseBuild
  tags: 
   - sql
  script: 
   - $projectPath = $env:CI_PROJECT_DIR
   - $project = "$projectPath\DoggosAreCoolDB\DoggosAreCoolDB.sqlproj"
   - $validatedProject = $project | Invoke-DatabaseBuild -TemporaryDatabaseServer "Data Source=PSE-LT-CHRISU\"

and we successfully built a database!

All that’s left to do now is 2 things:

  1. Create a NuGet package as part of the CI build
  2. Release the database changes to the target DB

I’m still using the same machine for the release portion too, so naturally I can use the same runner for this, if you have other servers you’re deploying to you will of course need additional runners.

We can very easily extend what we already have in our YAML file by just telling the process to create and export a new build artifact – I’m going to name it the same as everything else, and then append the BuildId to the end of the file so we always get something unique:

stages:
  - DatabaseBuild

Build:
  stage: DatabaseBuild
  tags: 
   - sql
  script: 
   - $projectPath = $env:CI_PROJECT_DIR
   - $project = "$projectPath\DoggosAreCoolDB\DoggosAreCoolDB.sqlproj"
   - $validatedProject = $project | Invoke-DatabaseBuild -TemporaryDatabaseServer "Data Source=PSE-LT-CHRISU\"
   - $buildArtifact = $validatedProject | New-DatabaseBuildArtifact -PackageId DoggosAreCool.Database -PackageVersion 1.$env:CI_JOB_ID
   - $buildArtifact | Export-DatabaseBuildArtifact -Path "$env:CI_PROJECT_DIR\Export"
  artifacts:
    paths:
     - $env:CI_PROJECT_DIR\Export\DoggosAreCool.Database.1.$env:CI_JOB_ID.nupkg
    expire_in: 1 week

You’ll notice how I’m exporting the NuGet package to the project directory and then uploading it, this is so that we’ll have access to it to release but also so that we can use the artifacts argument in our YAML to upload the file and make it a downloadable package through the GitLab interface (if you go to that SPECIFIC job):

Whilst we’re on a roll here (and things haven’t gone wrong for a while) I’m going to add 2 additional stages ALL AT ONCE to both “Create a Database Release Artifact” and “Deploy from a Database Release Artifact” using, once again, the SQL Change Automation PowerShell cmdlets.

Woo-Hoo! I’m invincible!

I broke it.

Can you see what I did wrong? The error is:

New-DatabaseReleaseArtifact : The specified value for the Source parameter is neither a valid
41database connection string nor a path to an existing NuGet package file or scripts folder:
42'CI_PROJECT_DIR\Export\DoggosAreCool.Database.1.725147351.nupkg'

So 2 fun things. 1 – I forgot to highlight there was an environment variable at one point, so it was just looking for the name of the variable in the path and 2) it keeps erroring out saying my NuGet file isn’t a NuGet file, weird right?

On further inspection it is yet another mistake I made. I’m using the job ID to name the NuGet package, which means when it tries to find the file it’s 2 steps ahead because each stage is counted as a different job! Duh!

sylvester stallone facepalm GIF

A few quick changes should hopefully sort this out! I’m going to put the instance of the pipeline ID in ($env:CI_PIPELINE_ID) and see if that makes a difference!

Wait. Did it just say the pipeline ran? SUCESSFULLY? That’s exactly what it said! We can verify that this actually happened as well by checking the DatabaseDeploymentResources folder for the Release Artifact to Integration:

And everything is there! Note you won’t have a changes.html report just yet because this is the first time we’ve successfully deployed to Integration, however if we run 1 more change through (I’ll add a stored procedure):

Boom.

Prince Harry Mic Drop GIF

Now of course we can add additional stages to this, for manual intervention or to promote to other environments, but I’m going to call it a win here and retire (until the next post) gracefully. I’m sure you’re all wondering what my final YAML file looked like too – well (counterintuitively) I’ve popped it all into GitHub for you and pasted it below. Enjoy!

stages:
  - DatabaseBuild
  - CreateRelease
  - DeployToIntegration

Build:
  stage: DatabaseBuild
  tags: 
   - sql
  script: 
   - $projectPath = $env:CI_PROJECT_DIR
   - $project = "$projectPath\DoggosAreCoolDB\DoggosAreCoolDB.sqlproj"
   - echo "Building project $project"
   - $validatedProject = $project | Invoke-DatabaseBuild -TemporaryDatabaseServer "Data Source=PSE-LT-CHRISU\"
   - $buildArtifact = $validatedProject | New-DatabaseBuildArtifact -PackageId DoggosAreCool.Database -PackageVersion 1.$env:CI_PIPELINE_ID
   - echo "Exporting artifact to $env:CI_PROJECT_DIR\Export"
   - $buildArtifact | Export-DatabaseBuildArtifact -Path "$env:CI_PROJECT_DIR\Export"
  artifacts:
    paths:
     - $env:CI_PROJECT_DIR\Export\DoggosAreCool.Database.1.$env:CI_PIPELINE_ID.nupkg
    expire_in: 1 week

CreateRelease:
  stage: CreateRelease
  tags: 
   - sql
  script: 
   - $integrationDB = New-DatabaseConnection -ServerInstance "PSE-LT-CHRISU\" -Database "BlogsDotRedgate_Integration"
   - $buildArtifact = "$env:CI_PROJECT_DIR\Export\DoggosAreCool.Database.1.$env:CI_PIPELINE_ID.nupkg"
   - echo "Creating Release Artifact for DoggosAreCuteDB - check C:\DatabaseDeploymentResources\DoggosAreCuteInc\ReleaseArtifacts\$env:CI_PIPELINE_ID\Integration for more information"
   - $releaseArtifact = New-DatabaseReleaseArtifact -Source $buildArtifact -Target $integrationDB
   - $releaseArtifact | Export-DatabaseReleaseArtifact -Path "C:\DatabaseDeploymentResources\DoggosAreCuteInc\ReleaseArtifacts\$env:CI_PIPELINE_ID\Integration" -Format Folder

Integration:
  stage: DeployToIntegration
  tags: 
   - sql
  script: 
   - $integrationDB = New-DatabaseConnection -ServerInstance "PSE-LT-CHRISU\" -Database "BlogsDotRedgate_Integration"
   - echo "Deploying changes to Integration"
   - Import-DatabaseReleaseArtifact -Path "C:\DatabaseDeploymentResources\DoggosAreCuteInc\ReleaseArtifacts\$env:CI_PIPELINE_ID\Integration" | Use-DatabaseReleaseArtifact -DeployTo $integrationDB

Delays are not DevOps, delaying DevOps is worse: Why we need better working practices now more than ever.

“The time is always right to do what is right.”
– Martin Luther King Jr.

Over the past few months, we have been on lock-down. The product of a devastating and deadly disease that has well and truly stamped it’s legacy on human history forever. But it is out of these times that we receive a glimpse, a look into what is possible, and what humanity can do. It is out of this fight, out of these ever decreasing odds that we finally see what a combined effort can do, and what a focus on our fellow people can bring about. Don’t know what I mean, take a look at the Good News Network and subscribe (just like me) to see the best of us.

But it only works, we only triumph, when we work together.

Recent Example: Scientists at Oxford University have seen a tremendous breakthrough with their virus efforts and are making unprecedented strides towards a viable vaccine – but it involves an incredibly strong partnership with pharmaceutical companies and governments worldwide for staged testing, large scale results gathering and continuous improvement.

So. Why then in the last few months have I been speaking to people who say things like:

  • We have delayed our non-BAU process implementation, because we have seen a spike in usage, and we need all hands on deck.
  • We have been investigating tooling and processes to help our development teams, but this is on hold as we come to terms with this newer way of working.
  • Our teams are keen to adopt more agile ways of working, but they are overburdened at the moment, so we have decided to postpone any research into this for the next few months.

It is feasible, amid a global panic, that people and companies will do (and definitely have done) what comes naturally to us; that is to “bunker down“. We believe that if we shift all efforts from projects and ongoing testing / new processes, we can have all hands available to deal with anything that comes our way. Processes are established for a reason, right? Legacy methods of dealing with ad-hoc changes and semi-frequent deployments, waterfall-esk development cycles and decade(s) old systems represent the familiar, the safe… Supposedly.

Now so, more than ever, it is time to actually change direction and to put more effort into some of the key principles and processes that will lead us to DevOps nirvana; it is this trinity of people, processes and tooling that can ultimately be the salvation for many global companies as they try to maintain their agility and competitiveness within an uncertain, shifting post-pandemic international market. There are many reasons why I, and many others, believe this but I have detailed 3 key reasons below:

1 – Delaying DevOps creates waste and costs businesses money

DevOps is a culmination of learning, experience and effort and it cannot be classified as one single thing, however it is possible to define a number of things that DevOps is and what it most certainly is not. One of the things that DevOps is, is “the constant delivery of value to end users“; the idea that by adopting certain technical measures and working practices we can minimize the time to delivery of new features and functions, which equates to greater value for us, our end users and significantly more agility to shift in different directions as required.

These ideas of flow and value streams are covered quite nicely by Lean IT, which extended from Lean Manufacturing principles and it is nicely explained here (and I would highly recommend you read The Phoenix Project if you haven’t already) – but the purpose of it (in a similar vein to agile principles) in this sense, is that it defines a number of things that don’t add value to the resulting product or service. These ‘things’ are referred to as waste and this is precisely what we should be looking to remove from our existing, legacy processes because why would we work on anything that doesn’t deliver any value?

There are a few different types of waste but I want to highlight three important ones that often exist as a result of legacy processes still being in place:

  • Defects Includes lack of testing (poor execution) and hot fixing environments (unauthorized changes)
  • Waiting Including everything from waiting for refreshed environments to waiting for feedback/results and even manual processes like deployment approvals
  • Motion (excess)Effectively doing the same thing over and over again, fire-fighting problems that arise on a near daily basis, engaging and monopolizing resources constantly who could otherwise be working on other, more important or value-add tickets (for those of you who HAVE read the Phoenix Project, see Brent as an example!)

The product of this waste is very simple and it fits into 3 main buckets: poor customer experience, increased costs and lost productivity. All of these things boil down to one fundamental truth – bad practices cost us money and reputation, transformation now could help us prevent this, and people will remember us for stepping up when we needed to.

2 – The workforce is increasingly distributed/remote and needs to collaborate better

At the beginning of 2020, one of the biggest questions faced by companies all around the globe was “how do I find and retain talent?”; this is not a new question and had already been around for years. Companies restricted to their offices (base OR satellite) realized they were increasingly fighting for candidates in one of the most competitive markets served by an ever dwindling local pool of options.

This was a situation which necessitated companies to stretch beyond their existing capabilities to enable a better quality experience for remote and distance workers and/or teams – a situation which would later be exacerbated 100x-fold by the global pandemic crisis. Not only has this crisis confirmed that most companies, certainly those feeding software markets, can work remotely, it has also posed the question of if we should work remotely more often, and has fundamentally changed the way we as a workforce will continue to work in the future.

One thing is for sure, whilst we will try to “return to normal” as much as possible, normal has been forever changed and remote working and collaboration is here to stay. That’s the important word in play here, collaboration.

The spirit of DevOps as I’ve mentioned before is good quality communication, collaboration and accountability. But at the heart of those three ideas is visibility. In an office we can over-hear, we can drop-in or bring things up “over the water cooler”/”at the coffee machine”. In a remote working context, that isn’t possible. So we have to adhere to 2 of the most important practices in modern day software development: transfer knowledge and record decisions.

When we work and communicate better in a remote/distributed workforce, and use tools and processes available to us, people don’t make unauthorized changes, or make decisions that affect you that you weren’t aware of until 3 weeks after they were made. It becomes easier to make decisions and generate better work faster, rather than being paralyzed by indecision and uncertainty as to whether you hold the latest version of the truth or if it is outdated. When we adopt the right processes and tools into our DevOps methodology, we know for certain what we should be doing, and why we should be doing it.

Process-wise this can easily take the form of common functions, many already at the disposal of teams when remote; stand-ups, retrospectives, mob- and pair- programming, OKR and sprint planning, there are lots of different ways for us to work well and know what we’re supposed to be working on at all times (and why). Tooling-wise we can then match these how we will be doing something with the respective record of what is being done, what decisions have been made and crucially, why. Using work management software like Trello, bug/feature tracking software like JIRA or Azure DevOps work items, source controlling everything (even having a strong branching and merging strategy to control workflow) with rigorous testing routines, policies and pull requests and automation all lead to better informed, happier*, well-performing developers with a crucial sense of purpose.

*Important side note: It is also crucial that we don’t simply lose ourselves in the business benefits completely – developers, testers etc. are all human and we all crave job satisfaction and happiness in our roles. Yes you might be able to increase your deployment frequency, minimize costs or complaints, but nothing compares to a satisfied, motivated team, which DevOps can help breed and inspire.

3 – DevOps breeds innovation and improves company performance, with a tangible return on investment (and not just financial!)

Automation is one of the single greatest ways we can modernize our processes, and is often the first principle we think of when adopting DevOps practices; taking something that is manual or held together by legacy scripts that forever fail and cause outages, and instead continuously integrating, continuously testing and continuously improving using the latest processes and tooling available to us. Automation allows us to create high cycle rates, enhance and multiply the feedback options we have within our pipeline(s) and allows us to reduce manual concerns and issues, to dedicate teams to the very thing they were employed to do: innovate. We only need watch how Netflix does DevOps to realize what we can unlock.

On a weekly basis I discuss existing processes with developers all over the world and one key trend always emerges that we need to focus on: they have a process that is currently manual, and it needs improving. I have lost track of the number of times I’ve been told about a process where developers generate scripts, manually test themselves and then “do x” with it, whether that be just deploying to Production themselves (without review), or putting it in an ever mounting pile of scripts in a folder on a file share for someone to sift through periodically.

Across every single one of the conversations mentioned above that I have, there is not a single discussion that doesn’t include some kind of quantifiable cost to the business, whether that be downtime, customer refunds, regulatory penalties and even in extreme cases, high developer churn. All can still be expressed in terms of 2 things: Time and Money.

If we remove those roadblocks for our developers, if we give them tools to enable them to more easily do their jobs, we put in place processes that allow them to more easily deliver that innovation, and tight automated controls to remove error-prone, manual jobs – we end up with something more akin to harmony. By this of course I mean the “constant delivery of value to end users“. This creates a more positive user experience, allows us to respond more quickly in an uncertain market and make decisions on what we should or, just as importantly, shouldn’t do, faster.

The world at the moment is a very uncertain place and has destroyed jobs, companies and whole industries. We should expect that consumer confidence is at an all time low and as we all come out of lock-down across the globe, we should be prepared to metaphorically “put our best foot forward” to help our developers believe in what we’re doing, stimulate faith in our industries and ultimately deliver more value bidirectionally.

But this idea only works if we do this now.

Conclusion

DevOps is more than just “picking up some new tools” or “rolling out agile” to development teams. It is fundamentally a mindset change that can drastically and fundamentally alter the underlying motivations and thinking within an organisation, allowing you to focus on the most important thing – delivering value, faster.

There are always times where delaying large scale roll-outs is a pertinent decision to make, and a hard one at that! But DevOps seeks to unify every part of of the development cycle; giving you greater visibility, communication, accountability and control, with maximum flexibility to test, validate and even pivot where needed.

But the time to do DevOps is not “once everything improves“, “once we get back to normal” or “when we have more time“… it is now. Now is the opportunity we all have to capitalize on the wave of change we have been consistently waiting to implement, to strengthen our position and future growth in our markets. DevOps is how we can come out on the other side of this disaster ready to embrace new technologies and ways of thinking, to respond to our customers needs, and deliver value and speed at scale.

3 methods for seeding test data during CI builds with Flyway

“It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.
Sir Arthur Conan Doyle, Sherlock Holmes

Can you tell I’m loving Flyway at the moment? Well I am. It’s JUST SO GOOD! Honestly there are so many things you can do with it! Don’t know what I’m talking about? Check out my posts on xRDBMS DevOps with Flyway and tSQLt unit tests with Flyway and you’ll see what I mean!

As a result of the above posts though I was asked a question that I had to think about for a little bit before having the best possible answer, how can we seed some testing data INTO the build database so that we can run some meaningful tests against it?

This makes perfect sense to me, but there’s also a few different ways to do this – so let’s go fly(way)!

flying i believe i can fly GIF

1 – Test Data Migration Scripts

In my previous posts on Flyway (above) I talked about having an entirely separate build folder present within the repository, and a folder of test migrations alongside our schema migrations – I called these the Build_Config folder, (containing the build configuration file) and the Test_Migrations folder (unsurprisingly containing testing migrations) in the _Migrations location:

I was using the same build config for 2 purposes; 1) to build the schema migrations from the base version, by passing it the Schema_Migrations location dynamically and 2) then building the tSQLt framework and testing objects by passing it the Test_Migrations location dynamically.

This actually worked surprisingly well, but even beyond this – the same method can be repurposed, or added to, by augmenting your testing scripts and adding a data insertion task (as an additional script or group of scripts). In my folder, I can simply add a migration like this:

Because of course I like dogs.

lana del rey yes GIF

and once pushed to the repository and the build has run we should be able to verify our testing data is present:

A bonus win for this step of course, is that where Devs have their own Flyway config files locally for their development databases they could also overwrite this behavior and point the testing and/or data scripts at their own database so they have some seed data to work with too!

2 – Add a data generation step to the pipeline

There are SO MANY technologies out on the inter-webs for generating data. SO MANY. Many of them also have a command line or PowerShell module that we can use to easily invoke them against a target, especially if that target is going to be persistent like my Flyway Azure SQL Build DBs!

Because I have access to it and because I’m using essentially SQL Server DBs, I could easily use Redgate SQL Data Generator – but to get the data you need you could use anything from DBATools Data Generation (also SQL Server) to FillDB for MySQL (which looks awesome and you could easily use this for Step 1 above too!)

There are numerous ways to invoke tools and applications and fortunately good CI/CD tools like Azure DevOps offer multiple ways to, for instance, run PowerShell or CLI steps from within the pipeline – so we could easily invoke SQL Data Generator on a VM or physical machine we have an Azure DevOps agent on – but this thinking also opens up the possibility of using something like Chocolatey to dynamically install the software on the Azure DevOps hosted pool VM during build (for the Redgate tools at the moment I suppose you’d need a Windows VM).

sassy pants chocolate GIF

I will be writing a future blog post about this step because it sounds _very_ interesting, but I’m not sure yet what can be done specifically using Chocolatey or if I’ll have to look elsewhere, although I have read this post in the past (thanks Paul!) detailing limitations and a great workaround using Azure DevOps, so it’s likely that’ll be my first port of call!

Just to give you an idea of end result with SQL Data Generator specifically though:

3 – Use existing data, don’t generate

Ok this one is going to be controversial already, I can tell! Let’s all stay calm!

happy chill GIF

The best data to be tested is our data. What we have in Production is what will have these changes deployed to it… eventually! So shouldn’t we just test against that? Well. Maybe, maybe not depending on what is in there.

There’s a few methods to achieve this – my personal favorite would be to use a SQL Clone, spin that up on a build VM rather than using an Azure SQL DB, and we can have all the data in an instant. Of course if we hold any sensitive PII/PHI then we should ensure that is protected first!

Of course there are lots of other options, like restoring a backup or spinning up a container etc. and these can all just be a stage in the YAML file before invoking Flyway but the point is, if we use an existing copy of our Prod database from some source or another, it will have 2 things we really care about:

  1. Data. Ready to go, ready to test, ready to give us the best possible insight into our changes.
  2. The flyway_schema_history table. Instead of running EVERY migration we’ve ever written, which could take a while for a large team, we run only the latest migrations to check that they would deploy happily to the Production target.

To get this stage to work though, you would need to do a couple of things differently:

  1. The build DB would have to be created from the clone/backup/other every time instead of simply cleaning the schema down.
  2. You would need to remove the Flyway Clean step from the pipeline in my previous post, because it would otherwise drop all the tables (and then we wouldn’t have any data!)
  3. By extension, this also makes the callback to remove the tSQLt objects void, so you can remove that too.

Conclusion

There are a lot of different ways to generate data, you can generate completely synthetic data, you can mask data or use Prod data, it’s up to you! Ultimately it will just for another part of your pipeline – just be careful of ordering! You don’t want to try generating data into a table that hasn’t been built yet.

Respect your YAML file and you’ll get schema, data and unit tests and this will lead to one thing. Greater insight, earlier.

thumbs up GIF

Flyway and tSQLt – migrating to warmer test climates

“If you truly have faith in your convictions, then your convictions should be able to stand criticism and testing.”
DaShanne Stokes

Welcome fellow TestDriven-Development enthusiasts… is what I would say if i actually ever did TDD and didn’t just, you know… write regular unit tests after the fact instead.

I’m going to be honest, I love the idea of TDD but have I ever actually been able to do it? No. Have competent developers been able to do it successfully? Yes, of course. Don’t know anything about TDD? You’re in luck! Click here for an introduction (don’t worry though, THIS post is not going to be about TDD anyway, so you can also keep reading).

But one thing we can all agree on is that testing is pretty important. Testing has evolved over the years though and there are a million-and-one ways to test your code, but one of the most difficult and frustrating things to test, from experience, is database code.

gilmore girls shot of cynicism GIF

Some people argue that the days of testing, indeed, the days of stored procedures themselves are gone and that everything we do in databases should be tested using a combination of different logic and scripting languages like Python or PowerShell… but we’re not quite there yet, are we?

Fortunately though we’re not alone in this endeavor, we have access to one of the best ways to test T-SQL code: tsqlt. You can read more about tsql at the site here but in short – we have WAYS to test your SQL Server* code. The only problem is, when you’re using a migrations approach… how?

*There are also many ways to unit test code from other RDBMS’ of course, like utPLSQL for Oracle Database or pgTAP for PostgreSQL – would this method work for those? Maybe! Try adapting the method below and let me know how you get on!

I’ve already talked about how implementing tests is easier for state based database source control in a previous post because we can easily filter tests out when deploying to later stage environments, however with migrations this can be a real pain because you have to effectively work on tests like you would any normal database changes, and maybe even check them in at the same time – so ultimately, they should be managed in the same way as database schema migrations… but we can’t filter them out of migrations or easily pick and choose what migrations get run against test and Prod, without a whole lot of manual intervention.

Basically. It’s a mess.

mess fail GIF

But during my last post about Flyway I was inspired. This simple and easy to use technology just seems to make things really easy and seemingly has an option for EVERYTHING, so the question I started asking myself was: “How hard would it be to adapt this pipeline to add unit tests?” and actually although there were complications, it was still easier than I thought it would be! Here’s how you can get up and running with the tSQLt framework and Flyway migrations.

1 – Download the scripts to create the tSQLt framework and tests from the site

Ok this was the easiest step of them all, largely because in the zip file you download from the tsqlt website all you have is a set of scripts, first needed to enable CLR and the second to install the tsqlt framework:

As part of my previous pipeline I’m actually using Azure SQL Database as my development environment, where RECONFIGURE is not a supported keyword and where we don’t need to run the CLR script anyway, so all I needed was the tSQLt.class.sql file.

The good thing about this is that we can copy it across into a migration and have this as our base test class migration, and then any tests we write on top of it will just extend it – so as long as we remember to update it _fairly_ frequently with any new tsqlt update, we should be fine! (Flyway won’t throw an error because these are non persistent build objects, so no awkward checksum violations to worry about!)

2 – Adapt the folder structure in the repository for tests

I added 2 new folders to my _Migrations top level folder, a Schema_Migrations folder and a Test_Migrations folder. When you pass Flyway a location for migrations, it will recursively scan folders in that location looking for migrations to run in order. I copied the migrations I had previously into the Schema Migrations folder and then my new tSQLt creating migration into the Test Migrations folder. This allows them to be easily coupled by developers, whether you’re writing unit tests or practicing TDD:

You’ll have noticed I called my base testing migration V900__ – this is because I do still want complete separation and if we have a V5 migration in schema migrations and a V5 testing migration, we’re going to have some problems.

3 – Add a callback to handle removal of the objects

As I was putting this together, I noticed that I could use flyway migrate to run the tSQLt framework against my Dev database, but every time I tried to then flyway clean that database I got a very nasty error stating that the tSQLt assembly could not be removed because of dependent objects.

Flyway does not handle complex dependencies very well unfortunately, that’s where you’d use an industry leading comparison tool like SQL Compare so, with some advise from teh wonderful Flyway team, I set to work on a callback. A callback is how you can hook into Flyway’s own processes, telling it to do something before, during or after certain commands. In my case we were going to remove all of the tSQLt objects prior to running Flyway clean to remove the rest of the schema. To make it future proof (in case objects are added or removed from the tSQLt framework), I wrote a couple of cursors to go through the different objects that were dependent on the assembly and remove them, rather than generating a script I know to have all of the tSQLt objects in right now. You can find the code for the callback in my GitHub here, you are welcome to it!

Animated GIF

All you have to do is name it beforeClean.sql and ensure it is in the directory with your other sql migrations so that it will pick this up and run it – I put it in my Test_Migrations folder, because I only want it to run this callback when cleaning the build DB, as this is the only place we’re utilizing automated unit tests… for now!

4 – Update the Azure DevOps pipeline

I’ve got my callback, I’ve got my tSQLt migration and the folder structure is all correct and is pushed to Azure DevOps but naturally it is breaking the build *sad* but fortunately all we now have to do is update the YAML pipeline file:

trigger:
- master

pool:
  vmImage: 'ubuntu-latest'

steps:
- task: DockerInstaller@0
  inputs:
    dockerVersion: '17.09.0-ce'
  displayName: 'Install Docker'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run -v $(FLYWAY_LOCATIONS)/Test_Migrations:/flyway/sql -v $(FLYWAY_CONFIG_FILES):/flyway/conf flyway/flyway clean -enterprise
  displayName: 'Clean build schema'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run -v $(FLYWAY_LOCATIONS)/Schema_Migrations:/flyway/sql -v $(FLYWAY_CONFIG_FILES):/flyway/conf flyway/flyway migrate -enterprise
  displayName: 'Run flyway for schema'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run -v $(FLYWAY_LOCATIONS)/Test_migrations:/flyway/sql -v $(FLYWAY_CONFIG_FILES):/flyway/conf flyway/flyway migrate -enterprise
  displayName: 'Run flyway for tSQLt'

You will notice a couple of important things that I have highlighted above:

  1. I’m cleaning the build schema using the Test_Migrations repository – this is because that is where my callback is and I need that to run before the clean otherwise it will fail due to the tSQLt assembly issue (line 17)
  2. I am running the migrate for the tests and the schema separately in the file, instead of just calling flyway to recursively run everything in the _Migrations folder. This is because I want them to be 2 separate steps, in case I need to modify or remove either one of them, or insert other steps in between and so that I can see the testing output in a separate stage of the CI pipeline (lines 23 and 29).

Caveat: As a result of (Option 2) running the 2 processes separately, it means running Flyway twice but specifying the Schema_Build and Test_Build folders in the YAML as being mapped to Flyway’s sql directory (lines 16 and 22 in the file above) but the problem this causes is that the second time Flyway runs, when it recursively scans the Test_Migrations folder it will not find the migrations that are present in the Flyway_Schema_History table, resulting in an error as Flyway is unable to find and resolve the migrations locally.

The way to fix this though is pretty simple – you find the line in the Flyway Config file that says “IgnoreMissingMigrations” which will allow it to easily continue. We wouldn’t have to worry about this setting though, if we were just recursively looking to migrate the Schema and Test migrations in the same step (but I’m a control freak tee-hee).

Now, once committed this all runs really successfully. Velvety smooth one might even say… but we’re not actually testing anything yet.

5 – Add some tests!

I’ve added a single tSQLt test to my repository (also available at the same GitHub link), it was originally created by George Mastros and is part of the SQLCop analysis tests – checking if I have any user procedures named “SP_”, as we know that is bad practice – and I have wrapped it up in a new tSQLt test class ready to run.

You’ll notice I also have a V999.9__ migration in the folder too, the purpose of this was to ‘top and tail’ the migrations; first have a script to set up tSQLt that could be easily maintained in isolation and then end with a script that lets me do just 1 thing: execute all of the tests. You can do this by simply executing:

EXEC tSQLt.RunAll

and we should be able to capture this output in the relevant stage of the pipeline.

Some of you may be asking why I chose to have the run unit tests as part of the setting up of the testing objects – this was because I had 2 options:

  1. I’m already executing scripts against the DB with Flyway, I may as well just carry on!
  2. The only other way I could think to do it was via a PowerShell script or run SQL job in Azure DevOps but the 2 plugins I tried fell over because I was using a Ubuntu machine for the build.

So naturally being the simple person I am, I opted for 1! But you could easily go for the second if you prefer!

6 – Test, Test, Test

Once you’ve handled the setup, got the callback in place (and also followed the steps from the last blog post to get this set up in the first place!) you should be able to commit it all these changes and have a build that runs, installs tSQLt and then runs your tests:

I realize there are a lot of “Warnings” in there, but that is just Azure DevOps capturing the output, the real part of this we’re interested in is lines 31-40 and if we clean up the warnings a little you’ll get:

+----------------------+
|Test Execution Summary|
+----------------------+
|No|Test Case Name|Dur(ms)|Result |
+--+---------------------------------------+-------+-------+ 
|1 |[somenewclass].[testProceduresNamedSP_]|144|Success|
------------------------------------------------------------
Test Case Summary: 
1 test case(s) executed, 1 succeeded, 0 failed, 0 errored. 
------------------------------------------------------------------

But if I introduce a migration to Flyway with a new Repeatable Migration that creates a stored procedure named SP_SomeNewProc…

+----------------------+
|Test Execution Summary|
+----------------------+
|No|Test Case Name|Dur(ms)|Result |
+--+---------------------------------------+-------+-------+ 
|1 |[somenewclass].[testProceduresNamedSP_]|184|Failure|
------------------------------------------------------------
Test Case Summary: 
1 test case(s) executed, 0 succeeded, 1 failed, 0 errored. 
------------------------------------------------------------------

It even tells us the name of the offending sproc:

All I have to do now is make the corresponding change to remove SP_ in dev against a bug fix branch, push it, create a PR, approve and merge it in and then boom, the build is right as rain again:

Thus bringing us back into line with standard acceptable practice, preventing us from delivering poor coding standards later in the pipeline and ensuring that we test our code before deploying.

Conclusion

Just because you adopt a more agile, migrations based method of database development and deployment, doesn’t mean that you have to give up on automated testing during Continuous Integration, and you can easily apply these same principles to any pipeline. With just a couple of tweaks you can easily have a fully automated Flyway pipeline (even xRDBMS) and incorporate Unit Tests too!

xRDBMS Database Continuous Integration with Flyway, Azure DevOps and Docker… the simple way.

“Some people try to make everything complicated, be the person who tries to make everything simple.”
Dave Waters

Simplicity is in my blood. That’s not to say I am ‘simple’ in the sense I cannot grasp more than the most basic concepts, but more that I am likely to grasp more complex problems and solutions when they are phrased in simple ways.

This stems from my love of teaching others (on the rare occasion it falls to me to do so), where I find the moment that everything just ‘clicks’ and the realization comes over them to be possibly one of the most satisfying moments one can enjoy in life.

shocked star trek GIF

Now recently I’ve been enjoying getting my head around Flyway – an open source JDBC based migrations tool that brings the power of schema versioning and deployments together with the agility that developers need to focus on innovation in Development. There’s something about Flyway that just… ‘clicks’.

It doesn’t really matter what relational database you’re using; MySQL, IBM DB2, even SAP HANA! You can achieve at least the core tenants of database DevOps with this neat and simple little command line tool – there’s not even an installer, you just have to unzip!

Now I’ve had a lot of fun working with Flyway so far and, thanks to a few people (Kendra, Julia – i’m looking at you both!) I have been able to wrap my head around it to, I would say, a fair standard. Caveat on that – being a pure SQL person please don’t ask me about Java based migrations, I’m not quite there yet!! But there is one thing that I kept asking myself:

“When I’m talking to colleagues and customers about Database DevOps, I’m always talking about the benefits of continuous integration; building the database from scratch to ensure that everything builds and validates…” etc. etc. so why haven’t I really come across this with Flyway yet?

think tom hanks GIF by The Late Show With Stephen Colbert

Probably for a few reasons. You can include Flyway as a plugin in your Maven and Gradle configurations, so people writing java projects already get that benefit. It can easily form part Flyway itself by virtue is simply small incremental scripts and developers can go backwards and forwards however and as many times as they like with the Flyway Migrate, Undo and Clean commands, so is there really a need for a build? And most importantly, Flyway’s API just allows you to build it in. So naturally you’re building WITH the application.

But naturally when you’re putting your code with other people’s code, things have to be tested and verified, and I like to do this in isolation too – especially for databases that are decoupled from the application, or if you have a number of micro-service style databases you’d want to test all in parallel etc. it’s a great way to shift left. So I started asking myself if there was some way I could implement a CI build using Flyway in Azure DevOps, like I would any of the other database tooling I use on a regular basis? Below you’ll find the product of my tinkering, and a whole heap of help from Julia and Kendra, without whom I would still be figuring out what Baseline does!

Option 1) The simplest option – cmdline

Flyway can be called via the command line and it doesn’t get more simple than that.

You can pass any number of arguments and switches to Flyways command line, including specifying what config files it’s going to be using – which means that all you have to do, is unzip the Flyway components on a dedicated build server (VM or on-prem) and then, after refreshing the migrations available, invoke the command line using Azure DevOps pipelines (or another CI tool) to run Flyway with the commands against a database on the build server (or somewhere accessible to the build server) and Bingo!

No Idea Build GIF by Rooster Teeth

And that’s all there is to it! You get to verify that all of the migrations up to the very latest in your VCS will run, and even if you don’t have the VERY base version as a baseline migration, you can still start with a copy of the database – you could even use a Clone for that!

But yes, this does require somewhere for Flyway to exist prior to us running with our migrations… wouldn’t it be even easier if we could do it without even having to unzip Flyway first?

Option 2) Also simple, but very cool! Flyway with Docker

Did you know that Flyway has it’s own docker image? No? Well it does!* Not only that but we can map our own version controlled Migration scripts and Config files to the container so that, if it can point at a database, you sure as heck know it’s going to migrate to it!

*Not sure what the heck all of this Docker/Container stuff is? You’re not alone! Check out this great video on all things containers from The Simple Engineer!

This was the method I tried, and it all started with putting a migration into Version Control. Much like I did for my post on using SQL Change Automation with Azure SQL DB – I set up a repo in Azure DevOps, cloned it down to my local machine and I added a folder for the migrations:

Into this I proceeded to add my base script for creating the DMDatabase (the database I use for EVERYTHING, for which you can find the scripts here):

Once I had included my migration I did the standard

Git add .
Git commit -m "Here is some code"
Git push

and I had a basis from which to work.

Next step then was making sure I had a database to work with. Now the beauty of Flyway means that it can easily support 20+ RDBMS’ so I was like a child at a candy store! I didn’t know what to pick!

For pure ease and again, simplicity, I went for good ol’ SQL Server – or to be precise, I created an Azure SQL Database (at the basic tier too so it’s only costing £3 per month!):

Now here’s where it gets customizable. You don’t NEED to actually even pass in a whole config file to this process. Because the Flyway container is going to spin up everything that would come with an install of Flyway, you can pass it switches to override the default behavior specified in the config file. You can adapt this either by hard-coding strings or by using Environment Variables alongside the native switches – this means you could pass in everything you might need securely through Azure Pipeline’s own methods.

I, on the other hand, was incredibly lazy and decided to use the same config file I use for my Dev environment, but I swapped out the JDBC connection to instead be my Build database:

I think saved this new conf file in my local repo under a folder named Build Configuration – in case I want to add any logic later on to include in the build (like the tSQLt framework and tests! Hint Hint!)

This means that I would only need to specify 2 things as variables, the location of my SQL migrations, and the config file. So the next challenge was getting the docker container up and running, which fortunately it’s very easy to do in Azure Pipelines, here was the entirety of the YAML to run Flyway in a container (and do nothing with it yet):

trigger:
- master

pool:
  vmImage: 'ubuntu-latest'

steps:
- task: DockerInstaller@0
  inputs:
    dockerVersion: '17.09.0-ce'
  displayName: 'Install Docker'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run flyway/flyway -v
  displayName: 'Run Flyway'

So, on any changes to the main branch we’ll be spinning up a Linux VM, grabbing Docker and firing up the Flyway container. That’s it. Simple.

So now I just have to pass in my config file, which is already in my ‘build config’ folder, and my migrations which are in my VCS root. To do this it was a case of mapping where Azure DevOps stores the files from Git during the build to the containers own mount location in which it expects to find the relevant conf and sql files. Fortunately Flyway and Docker have some pretty snazzy and super clear documentation on this – so it was a case of using:

-v [my sql files in vcs]:/flyway/sql

as part of the run – though I had to ensure I also cleaned the build environment first, otherwise it would just be like deploying to a regular database, and we want to make sure we can build from the ground up every single time! This lead to me having the following environment variables:

As, rather helpfully, all of our files from Git are copied to the working directory during the build and we can use the environment variable $(Build.Repository.LocalPath) to grab them! This lead to me updating my YAML to actually do some Flyway running when we spin up the container!

trigger:
- master

pool:
  vmImage: 'ubuntu-latest'

steps:
- task: DockerInstaller@0
  inputs:
    dockerVersion: '17.09.0-ce'
  displayName: 'Install Docker'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run -v $(FLYWAY_LOCATIONS):/flyway/sql -v $(FLYWAY_CONFIG_FILES):/flyway/conf flyway/flyway clean -enterprise
  displayName: 'Clean build schema'

- task: Bash@3
  inputs:
    targettype: 'inline'
    script: docker run -v $(FLYWAY_LOCATIONS):/flyway/sql -v $(FLYWAY_CONFIG_FILES):/flyway/conf flyway/flyway migrate -enterprise
  displayName: 'Run flyway for schema'

Effectively, this will spin up the VM in ADO, download and install Docker, fire up the Flyway container and then 1) clean the target schema (my Azure SQL DB in this case) and 2) then migrate all of the migrations scripts in the repo up to the latest version – and this all seemed to work great!*

*Note: I have an enterprise Flyway licenses which enables loads of great features and support, different version comparisons can be found described here.

So now, whenever I add Flyway SQL migrations to my repo as part of a branch, I can create a PR, merge them back into Trunk and trigger an automatic build against my Flyway build DB in Azure SQL:

Conclusion

Getting up and running with Flyway is so very very easy, anyone can do it – it’s part of the beauty of the technology, but it turns out getting the build up and running too, when you’re not just embedding it directly within your application, is just as straightforward and it was a great learning curve for me!

The best part about this though – is that everything above can be achieved using pretty much any relational database management system you would like, either via the command line and a dedicated build server, or via the Docker container at build time. So get building!

ready lets go GIF