This is about some interesting behaviour around workflows & managed solutions, which I’ve recently discovered. Let me give a bit of background first.

Currently I’m working on several COVID-19 apps for local authorities, to be able to help them assist people in need. As part of this, each local authority has a portal within the solution. The portal itself is a Power App Portal, and I haven’t really had exposure to them before.

Installing a Power Apps Portal comes with quite a large number of solutions in order to get it to work. More on this below.

Due to the way in which we’re engaging with our clients, the solutions are built in a single tenancy (different environments, of course!). We’re then inviting the users in as guests through Azure Active Directory, to be able to access functionality etc. This works well – we don’t need to worry about managing user accounts, AAD permissions, etc. However it also means that we don’t have any Office 365 licenses within the environment itself.

Now we have workflows that are sending emails out around the portal – registrations, password resets, etc. These are being generated automatically by the system, but as there’s no Office 365 mailbox for the user, they’re queuing up.

It’s not possible to authenticate a mailbox belonging to an external user (we tried!), as the system needs a native (full) user with an active mailbox to be able to send out emails. This is of course unlike Power Automate, where you can create a Send Email action and use specified credentials for logging in to send an email.

So, we did what any normal system administrator/configurator would do. We opened up the relevant (managed) solution, and from there opened up the workflow that we needed to modify. Things looked normal at first – we deactivated the workflow, and started poking around it to see what made it tick.

We came across the part that actually took user credentials to send the email that was being generated, and modified this accordingly. Then we saved the workflow, which was successful. However, upon trying to then reactivate the workflow, we got the following error message (helpful, isn’t it!):

Nicely it gives the option to download the log file around the error. This can usually be quite helpful (at times), so we thought we’d take a look at it. Behold the following (I’ve had to shrink the screenshot to allow it to fit on the screen!):

Isn’t that ‘beautiful’. Don’t worry if you can’t actually make out the error information – none of it makes any sense, at least not in a practical sort of way.

Being stuck at this, I thought to reach out to one of the community Power App Portal champions, Mario Trueba. I’ve known him for a while, and he’s just simply amazing. Having asked if I could jump on a call with him for 15 minutes to diagnose (& hopefully find an answer!), we spent almost an hour!

He suggested trying to use the classic interface, as I had been doing all of this through the new UI. So off I went to open up Classic (I’ve missed this, I will freely admit). Through there, we opened up the solution, opened up the workflow, and re-activated it. Or not, as it happens – even through the Classic UI, we weren’t able to do so. We tried a variety of things, but to no avail. It just simply wasn’t happening!

I was slightly concerned that there was an underlying issues with Portals, perhaps from some legacy CafeX code. I had tried searching with Mario for error details contained within the log file, but we couldn’t find anything that would fix it.

The next morning on waking up & checking Twitter, I noticed someone tweeting around Portals, and engaged with them. They turned out to be on the Portals development team, and told me to shoot them over an email with the details, which I did. They then replied to me, saying that it wasn’t anything specific to Portals, and that I should raise a support ticket. That crossed one item off my list (a Portals issue), but I was still needing to get things resolved.

So I went off & raised a support ticket. A few hours later, a very nice tech support person called Siva gave me a call to discuss the issue. We hopped into Teams, and in what I can only describe as the SHORTEST period of time that I’ve ever experienced, the issue was resolved (it took 7 minutes in total. Yes, I know…). Don’t worry – I’m not going to leave you hanging here!

See, what the ‘issue’ (and I’m deliberately putting it in quotes) was turned out to be something quite simple, yet quite strange.

Essentially opening the workflow from the managed solution somehow (& I don’t know HOW) inherits the ‘managed’ property. This is whether we open it from the new UI, or the classic UI. As a result we’re able to deactivate it, but we CAN’T reactivate it due to the system thinking that we’re modifying a managed component (as an aside, it is interesting how I did manage to save it though?). This was what was causing things to fall over, and the error message was really not helpful at all.

It’s also not a matter of being a Microsoft (or ISV) managed solution. I’ve replicated this happening with a solution that I’ve built, exported as managed, & then imported.

So how did we do it? Well, there are two ways in which this can be dealt with:

Either we can go to System/Processes, find the workflow there, open it up, and then reactivate it:

Or we can open up the Default solution, navigate to processes, select the workflow, and then reactivate it:

Both methods work just fine, and as mentioned earlier on, I’ve since replicated this on workflows in other managed solutions.

To me, this is somewhat strange, and should work regardless. According to Siva, it’s the desired system behaviour, though I have no idea why someone should want it to work in one way, and not in another.

So if you’re reading this, and you might just happen to know someone in the necessary Microsoft engineering/development team who’d be able to answer this, could you point them my way? I’d love to engage them to find out why, how, and if they could pretty please change this?