Tag: deployment

Metadata behaviour with solutions

Let’s start with a definition:

Metadata is a set of data that describes and gives information about other data

Crystal clear, right? Well, to anyone who’s unfamiliar with the concept of metadata, this is probably NOT so clear.

So what is metadata? Well, the way that I like to describe it is as a set of data that contains information about something. So for example, if I were describing a car, the metadata could be:

  • Manufacterer
  • Model
  • Colour
  • Engine size
  • etc

When it comes to Power Platform, the metadata is referring to the solution structure, ie tables, columns (& column types), etc. Essentially it describes the solution, but doesn’t hold any data that’s entered by users. In fact, we could describe the entire solution file as metadata, which we build & then deploy to downstream environments!

However, the solution metadata isn’t just describing the tables included in the solution, the columns etc. It also describes (& defines) the actual attributes, such as the table type, whether duplicate detection rules are activated or not, etc.

Now, when creating a new table, the metadata is created automatically for us – we don’t need to do anything about it (well, apart from setting it as we may wish it to behave). When importing a table into a solution, however, we get the lovely option to include metadata or not:

As we’re going to see, this is actually quite an important little item to be aware of.

See, if we don’t include the table metadata, we can still set the table properties within the solution. So what is the problem then? Well, there is actually an underlying challenge, which we found out about this week.

We were developing a solution for a customer, which included some standard tables (eg account, contact, etc). We added these into our (custom) solution, though on adding, we did NOT tick the ‘Include table metadata’ option.

We went ahead with creating our custom columns, forms, views etc, all normal sort of practise. However, one of the business requirements stated that auditing needed to be present. Well, there’s a standard table audit functionality available in Power Platform, so we went ahead & selected this in our custom solution:

OK – all is looking good. We finished with the development, and started to go through the ALM process.

On importing the solution to UAT, we couldn’t see any auditing going on (yes, we made sure to enable auditing in the environment settings!). Looking at the solution in UAT, the table did NOT have the auditing flag set for it.

This was quite puzzling. We went back to the DEV environment, and sure enough, the auditing flag was set there. However it obviously wasn’t going through to the UAT environment.

Opening up the actual solution file exported from DEV gave us a surprise. Though we had set the auditing value in it, it wasn’t actually included in the XML:

This was quite puzzling. We decided to try again from the begining with adding the table to the solution, but also setting the ‘Include table metadata’ flag when doing so. We went through the process again of exporting the solution and opening up the file to inspect it. This time, the table audit value WAS present (along with a lot of other metadata values as well):

This was, to be honest, a little annoying. We specifically didn’t want to bring in all the table metadata, as we could potentially affect another setting.

Reaching out to Microsoft, we sought clarification on the behaviour of this. They very nicely came back to provide further information.

At the moment, the behaviour is such that if a user doesn’t include the table metadata when adding it to a solution, it doesn’t matter if the user sets any attributes/values within the solution.

There is a single flag to say whether or not all metadata of a table should be exported or skipped when a solution is exported. This will include ALL includes all attributes ‘about’ the table – display name, description, audit flags, etc. It is quite extensive (for more information on all of the metadata options, please visit EntityMetadata EntityType (Microsoft.Dynamics.CRM) | Microsoft Learn)

If we have a need to modify/change metadata behaviour, we have to set the ‘Include table metadata’ flag. At this point in time, it’s NOT possible to select which metadata we want to include – it’s all or nothing. Now, that’s not to say that this may not change in the future – I can definitely see a need for this, and have fed this back into the Microsoft team accordingly.

I’m hoping that this will come about at some point to allow selective metadata inclusion within the solution, as I think it will be of great benefit to developers!

Has this scenario ever tripped you up before? Do you have scenarios that having selective metadata options would benefit you? Drop a comment below – I’d love to hear!

Solution deployments: Automated vs Manual

Over the holiday period, I’ve been playing around with solution deployments. OK – don’t judge me too much…I also took the necessary time off to relax & get time off work!

But with some spare time in the evenings, I decided to look a bit deeper into the world of DevOps (more specifically, Azure DevOps), and how it works. I’ll admit that I did have some ulterior motives around it (for a project that I’m working on), but it was good to be able to get some time to do this.

So why am I writing this post? Well, there’s a variety of great material out there already around DevOps, such as https://benediktbergmann.eu/ by Benedikt (check out his Twitter here), who’s really great at this. I chat to him from time to time around DevOps, to be able to understand it better.

However, I ran into some quite interesting behaviour (which I STILL have no idea why it’s the case, but more on this later), and thought that I would document it.

Right – let’s start off with manual deployments. As we know, manual deployments are done through the user interface. A user (with necessary permissions) would do the following:

  1. Go into the DEV environment, and export the solution (regardless of whether this is managed or unmanaged)
  2. Go into the target environment, and import the solution

Pretty simple, right?

Now, from an DevOps point of view, the process is similar, though not quite the same. Let’s see how it works:

  1. Run a Build pipeline, which will export the solution from the DEV environment, and put it into the repository
  2. Run a Release pipeline, which will get the solution from the repository, and deploy it to the necessary environment/s

All of that runs (usually) quite smoothly, which is great.

Now, let’s talk for a minute about managed solutions. I’m not going to get into the (heated) discussion around managed vs unmanaged solutions. There’s enough that’s been written, said, and debated on around the topic to date, and I’m sure it will continue. Obviously we all know that the Microsoft Best Practise approach is to use managed solutions in all non-DEV environments..

Anyway – why am I bringing this up? Well, there’s one key different in behaviour when deploying a managed solution vs an unmanaged solution (for a newer solution version), and this is to do with removing functionality from the solution in the DEV environment:

  • When deploying an unmanaged solution, it’s possible to remove items from the solution in the DEV environment, but when deploying to other environments, those items will still remain, even though they’re not present in the solution. Unmanaged solution deployments are additive only, and will not not remove any components
  • When deploying a managed solution, any items removed from the solution in the DEV environment, and then deploying the solution to other environments will cause those items to be removed from there as well. Managed deployments are both additive & subtractive (ie if a component isn’t present in the solution, it will remove it when the solution is deployed)

Now most of us know this already, which is great. It’s a very useful way to handle matters, and can assist with handling a variety of scenarios.

So, let’s go back to my first question – why am I writing this post? Well..it’s because of the different behaviour in manual vs automated deployment, which I discovered. Let’s look at this.

When deploying manually, we get the following options:

The default behaviour (outlined above) is to UPGRADE the solution. This will apply the solution with both additive & detractive behaviour. This is what we’re generally used to, and essentially the behaviour that we’d expect with a managed solution.

Now, when running a release pipeline from Azure DevOps, we’d expect this to work in the same way. After all, systems should be build to all work in the same way, right?

Well, no, that’s not actually what happens. See, when an Azure DevOps release pipeline runs, the default behaviour is NOT to import the solution (we’re talking managed solutions here) as an upgrade. Instead (by default), it imports it as an UPDATE!!!

This is what was really confusing me. I had removed functionality in DEV, ran the build pipeline, then ran the release pipeline. However the functionality (which I had removed from DEV) was still present in UAT! It took me a while to find out what was actually happening underneath…

So how can we handle this? Well, apart from suggesting to Microsoft that they should (perhaps) make everything work in the SAME way, there’s a way to handle it within the release pipeline. For this, it’s necessary to do two things:

Firstly, on the ‘Import Solution’ task, we need to set it to import as a holding solution.

Secondly, we then need to use the ‘Apply Solution Upgrade’ task in the release pipeline

What this will do is then upgrade the existing solution in the target environment with the holding solution that’s just been deployed.

Note: You will need to change the solution version to a higher solution number, in order for this to work properly. I’m going to write more about this another time, but it is important to know!

So in my view, this is a bit annoying, and perhaps Microsoft will change the default behaviour within DevOps at some point. But for the moment, it’s necessary to do.

Has this (or something similar) tripped you up in the past? How did you figure it out? Drop a comment below – I’d love to hear!

Workaround for sharing Canvas Apps

Don’t you find it absolutely frustrating when there’s a canvas app that you want to get access to, or give other users access to, but can’t see it? It’s REALLY annoying, but it’s sort of the way that Microsoft has designed the platform (at least at this point in time).

See, when a user creates a canvas app, only the creator is able to see & launch it. If other users want to get access to it, the creator needs to share it. This can be done by sharing the app directly with another user, or by sharing it with an AAD Security Group (which is sort of best practise).

Now, of course there’s the Microsoft Power Platform Centre of Excellence solution, which includes a very handy app to assign permissions for canvas apps. After all, if a user is on holiday, sick leave, or has left the company, there needs to be some way of assigning permissions for other users to gain access to it. It’s really helpful, but of course needs the CoE solution installed.

Let’s think of another scenario. What about if we have some canvas apps as part of a solution, that’s deployed through (proper) ALM – such as using Azure DevOps with automated pipelines. Best practise for this is to use service principals (ie non-interactive user logins). This is great, but then the canvas app/s will be owned by this user. So without the use of the CoE ‘Set App Permissions’ canvas app, we’re sort of stuck, as we can’t gain access to the app.

Or can we…..?

So this is a scenario that I’ve been dealing with recently, and I’ve found a really cool workaround that doesn’t need the CoE ‘Set App Permissions’ canvas app to be able to handle the situation.

The example below (amusingly, in my opinion) is actually using the Microsoft CoE solution as an example, but this works with any canvas apps that are held within a solution (against, this heavily supports using solutions for ALL development items!).

So, this is what the actual installed apps look like in this environment:

As we can see, there are a lot of them! But what happens if I’m logged in as my regular user? What do I see if I go to the list of apps? Well, I’ll see the following:

Now, as we can see, I’m able to see the model-driven app (as these aren’t hidden at all). But I’m not able to see ANY of the canvas app! So how can I get access to it, or share it with other users?

Well, if I take a look at the solution itself, I can see the following when browsing to the list of apps (I’m really loving the new Solution Explorer layout, I’ll freely admit!):

I can try to play the canvas app (in this case, the ‘Set App Permissions’ app) directly from the solution. But when I try to do this, I’ll get the following error message:

Now, this is of course happening because I’m not the owner of the app, & the app hasn’t been shared with me at all. So really I was expecting this error to happen.

However, if I take a look at the menu options displayed for me, I can see that the ‘Share’ option isn’t greyed out. I wonder what happens if I click it…

Now this is EXCITING! When clicking the ‘Share’ option on the menu, I’m given the regular sharing screen, where I’m able to set app permissions. So it looks like I’m able to do something here. OK – let’s go ahead & try to share the app with my own user:

So I’ve looked up my own user, and then clicked ‘Share’. This is what happens next…

Exciting moment – will this work?

Waiting with bated breath, and then…

It’s worked! The app sharing has been successful with my user.

Note: The example that I’m using here is with my own user account. However it doesn’t need to be – I can select any user account or AAD Security Group, and share accordingly.

Going to my list of apps, I can now see that the app is showing up for me:

Clicking the app to launch it presents me with the permissions dialogue, and having confirmed permissions, then launches it properly:

So this is indeed a way in which it’s possible to share canvas apps with users and/or AAD security groups, even when a user isn’t the owner of the canvas app.

It is important to note that the user carrying this out does need to have one of the following permissions in the environment:

  • System Customiser
  • System Administrator

Without having one of these roles, it’s not going to be possible to carry out the above (mostly because it’s not possible to see solutions & dig down into them).

This is a handy little trick that hopefully will help clear up one of the headaches when trying to share canvas apps! Of course it’s possible to use the Microsoft CoE tool to set app permissions, but if a customer doesn’t have it installed, then this would be another way to approach things.

Have you ever had this issue? How did you go about solving it? I’d love to hear – please drop a comment below…

Canvas Apps & Power Automates

So it’s been a busy few weeks here, which is why I haven’t really been putting up any articles. March/April is always a busy time for our family with stuff going on, and this year I decided not to push myself to get articles out, as otherwise I’d be running very low on sleep!

That being said, I’ve still had some great ideas about things that I’d like to share, and have been keeping a series of short notes for me to pick up. Today’s topic is one of them, which I think has been a major pain to anyone involved in canvas app development!

So, the back story to this is that we’re able to use Power Automate flows together with canvas apps. What I mean by this is that we’re able to directly trigger them from within the canvas app, rather than needing to do something like edit or create a record, and then have the Power Automate flow trigger from the record creation or modification.

There’s a specific Power Apps trigger that’s available within Power Automate exactly for this purpose:

When clicked, it gives us the trigger line in the steps as follows:

So what we’d do is within the canvas app, we would bind a button (or another control) that when selected, it would then go away & trigger the Power Automate flow. Great – so many different things that we can get to happen! One of the benefits of doing things like this is that we can then pass information from the Power Automate flow back to the canvas app directly:

This can then mean that the user can know, within the canvas app itself, that the Power Automate flow has run, and use data (or other things) that have come out of it.

OK – all good so far.

The main issue to date has been with deploying canvas apps together with Power Automate flows. See, as per best practise, we would create a solution, place the canvas app, flows, and anything else that’s necessary for it to work within it, and then deploy the solution to our target environment/s. And that’s where things just…didn’t go quite right.

Obviously within the development environment, the canvas app would be hooked up to the flows, and everything would work. Clicking the button would cause the flow to run, etc. User authentication would be in place (along with licenses of course!), and it was just fine.

But when deploying a solution containing canvas apps and associated flows between environments (regardless of whether it’s been manually deploying, or automated using a tool such as Azure DevOps), the connections to the flows would be broken. Ie, the canvas app would run, but the flows wouldn’t trigger. Looking at the connections in the canvas app within Studio would show something like the following:

All of the connections to Power Automate flows would show as ‘Not connected’. It’s not even possible to click the ellipse next to them and re-connect them – the only option available is to remove it from the canvas app!

So in order to get things working again, we’d need to do the following steps:

  • Open up the canvas app
  • Remove all connections to Power Automate flows
  • Add a temporary button, set it to be a Power Automate trigger
  • Click through all of the Power Automates needing to be connected (waiting for each one to connect, then go to the next one)
  • Remove the temporary button
  • Save and publish the solution

This, in a nutshell, has been a (major) headache. For example, I’ve been working with a solution that has over 30 Power Automate flows that can be triggered from the canvas app (lots of different functionality!). Each deployment has needed the above process to be carried out, which has usually added on at least an hour to the deployment process!

Now, this hasn’t been something that’s been unknown. In fact, the official Microsoft documentation noted the following:

So this is something that Microsoft has been well aware of, but it’s been a pain point that we’ve had to work with.

However, this has now ALL changed, which I (and MANY others) are really pleased about!

Microsoft has rolled out an update last month that means that canvas app connections to Power Automate flows will NOT break when they’re deployed across environments! This is such a massive time-saver, that I’m now trying to work out what to do with all of my free time! Only kidding…more project work will commence!

So what we can now do is take our solution, deploy it across the different environment/s that we need to get it out to (whether manually, or automated using tools such as Azure DevOps), publish the solution, and then everything works! Amazing!!

One small caveat though – to ensure that this work, you will need to go into the app, and re-publish it on the latest Power Apps version. This should of course be done in a development environment, and then can be exported and deployed as required.

Microsoft have also updated their documentation at https://docs.microsoft.com/en-us/powerapps/maker/data-platform/solutions-overview to remove the limitation text shown above. It’s a good place to keep an eye on changes that occur over time too.

This is definitely a welcome piece of development, and I know that we’ve been eagerly waiting for this for a while, and now it’s here!

‘Ghost’ lookup value following deployment

This is something that stumped me fairly recently. It’s also something that I was trying to work out what I should use at the title for this post! Let me share what happened.

I’m working on a project that’s quite critical (COVID-19 related). This is a project that we’ve built something around Dynamics 365 as an additional wrapper, to provide specific functionality for the pandemic. It’s being rolled out (the same solution) to multiple clients, and is only using the functionality from Power Platform. No custom code at all.

Now, before going into the specifics around it, let’s take a moment to revisit what a lookup field is, and what it does. Essentially a lookup field connects two tables together (wow – that felt strange not to use the word ‘entity’!). In the front interface, it’s used for a 1:N relationship.

So for example, we can have a lookup from Account to Contact, to set the primary contact for the account. The user navigates to the field, searches for the record they’re wanting to associate, and saves it.

Underneath, there’s a relationship that’s automatically created between the two tables, showing the way that the relationship will go (ie 1:N or N:1). This is created on both sides (more on that another time around dependencies), and most people will never need to modify it

When I first started with this particular project, I got the solution, and deployed it into the Dev environment (for the project that I was on). On testing it out, I found something very interesting. We’re using the Case (Incident) table, and there are various lookup fields on it. One of these was already populated with a value. Hmm – that’s interesting, I thought. It was a new deployment, and we hadn’t set any static data up yet at all. So how could it already be populated?

How is this being set, when I’ve not entered it into the system as a record…

Furthermore, I was unable to save the Case record. When I tried to, I was getting an interesting error:

On drilling down into the error log (which admittedly is actually getting better in the details shown in it, thankfully!), it turned out to be because I didn’t have access to the referenced record (in the lookup field). It just didn’t exist.

So the lookup field value was coming in with a hard-coded GUID (record identifier). But how was this being done, especially if there weren’t any records (of that type) in the system at all?

From my experience of things, I could think of two ways in which to populate a lookup field with a hard-coded value:

  • Through a ‘real-time’ Power Automate flow, on create of the record. It’s possible to set a GUID value in the flow, and then it would be set
  • Through custom code, running on the form. Again, it’s possible to hard-code a GUID there, and then set the field

However on checking both options, none of them were happening. No Power Automate flows touching the Case record, and no custom code at all on the Case.

It was then, digging through the other parts of the solution, that I saw various Business Rules. For those unfamiliar with these, I’ll quote from the official Microsoft documentation around them:

By combining conditions and actions, you can do any of the following with business rules:

  • Set column values
  • Clear column values
  • Set column requirement levels
  • Show or hide columns
  • Enable or disable columns
  • Validate data and show error messages
  • Create business recommendations based on business intelligence.

I’ve used Business Rules (somewhat extensively) before. However on going into the one for the Case table, I found that something was happening that I wasn’t aware could happen! It’s actually possible to set a lookup field value through it:

I spy a lookup option

Even though we’ve deployed the solution from the original development environment to a different environment, this is still set. But there are no records that are available:

I had never thought that it would be possible – to set a static value (eg a number, or some text), fine. But to set referential data? Wow.

Obviously this can be quite helpful. The bit that it’s NOT helpful though is when deploying the solution to another environment (as this situation was). It doesn’t help if you re-create the record that it’s referring to with using the same record name, as it’s using the underlying GUID (which you can’t re-create). This really does take solution deployment into a whole new perspective, where you need to be careful around these sorts of things as well.

So something new that I’ve learned (I do try to learn something new each day), and specifically around an area I thought I knew quite well. It did take some time, but I’m glad that I (finally) found the root cause of it, and identified what was causing it.

Have you ever had something like this happen, where you’re searching & searching for the cause of it? Drop a line below – I’d love to hear!