Which Agile methodology to use during ERP implementations
Agile SCRUM vs KANBAN
Do not use Agile SCRUM on customer implementations. With this one liner we start this blog. Why not and what should you use then?
In general, we see these days that all ERP implementations are done in an Agile way; the waterfall methodology has become an exception. But there are various Agile methodologies, the most common ones are SCRUM and KANBAN.
Customer projects
The method to use is important at Customer projects in our role of implementation partner (ERP/CRM implementations) but also on the ISV products we maintain ourselves as Kaya. Leading factor to determine which methodology to pick is not the methodology itself but how an organization is working and organized. Beside that the necessary skillset for building and releasing software plays a role too. Those skills are usually less present at a customer compared to a Microsoft, or other software, partner.
ISV products
In case you are in full control of your resources & requirements, planning fixed (release) dates is working fine. We see this at the implementation of ISV products. So, for those using Agile SCRUM would be a good choice.
However, on Customer implementations, we as a partner are not always in full control. Some examples are:
- Decisions that requirements are out of scope (read not before go-live)
- Capacity of customer resources are hard to plan. A partner has no control over it.
- Customer must accept the solution and usually they need support on the testing as they are not used to this.
So, the planning of a customer implementation needs much more flexibility. This flexibility can be better arranged with KANBAN. To have a closer look at what KANBAN is use following link https://www.scrum.org/resources/kanban-guide-scrum-teams.
Differences between Kanban and Scrum
Most important differences between KANBAN and SCRUM are:
| Scrum | Kanban | |
| Cadence Regular | fixed length sprints | Continuous flow |
| Release methodology | At the end of each sprint | Continuous delivery |
| Roles | Product owner, scrum master, development team | Stakeholder |
| Key metrics | Velocity | ead time, cycle time, WIP |
| Change philosoph. | Avoid changes during the sprint. | Change can happen |
Let us start with the Cadence. The functional confirmation of D365 F&S has a higher throughput than the development of code extensions. Beside the code extensions we also have the ISV solutions and the MS monthly updates.
This brings me to the next point. Release mythology: a code release is different from replicating a setup in a target environment.
Then we have the Roles. There are several Scrum Roles, where customers usually are not accustomed to, and the theory is that Kanban has no roles. To my opinion, this is not the case there is a stakeholder, wo defines what he needs and signs of when his wishes are realized. To a certain extend this resembles the Product owner role.
The change philosophy of KANBAN is more flexible. In case you have not started on a card, you can always put it back in the backlog. We see this happen when the Implementation gets disrupted by day to day topics or customer changes scope or priority.
The impact of using KANBAN at a customer implementation has an additional dimension. This dimension is the division between Development and Operations. In general, we can say the Partner is development and the customer is operations. This helps to determine who is responsible for what.
The next diagram narrows it even more clear on extensions to be delivered during implementation. The Partner (DEV) delivers a combined build into the asset library, it is up to the Customer (OPS) to continue from that point onwards.
| OPS | DEV | DEV | OPS | OPS |
| Customer | Partner | Customer | Customer | |
| Backlog | DEV | TEST | Acceptance | Production |
| Code changes | Code changes | LCS asset | LCS asset | |
| Manual setup | Manual setup | Manual Setup |
Additions to Kanban
Beside the concept of KANBAN and DEVOPS, Kaya also uses the phrase requested today, delivered tomorrow.
- When a Developer is finished, the solution is automatically available next day in TEST
- When a Tester is finished, he has in general a release candidate feature for the Customer
- The customer can almost do cherry picking on the release candidate features. The ones the customer select will be available next day in the combined build in the LCS asset library.
Are we in control?
So, are we in control? Yes! But be careful there is still a risk. This is so called work in Progress (WIP). The WIP on Acceptance is extremely important, all acceptance task must be completed before going to production. Who is responsible? The Customer (stakeholder) is responsible, but of course supported by the partner (which task should be completed first, dependencies, risks etc.). How can the Customer control it? Simply throttle on the requests that are picked from the backlog.
For other DEVOPS tips have a look at https://kaya-consulting.com/category/lcs/
Can you sell me Regression Suite Automation Tool (RSAT)?
With the current fast appearance of new versions of D365 FO (10 times a year). Testing can become more time consuming. So, you must decide:
- Skip updates, take for example only the even or odd platform update numbers
- Do not test and have confidence at MS. But please be aware at MS they are also mere humans and humans make mistakes.
- Increase your Test team so more tests can be performed in a shorter time.
- Start using the Regression Suite Automation Tool.
In the past we chose option 1. We did still release every month, but the release of MS was not included the releases with ISV and tailored solutions for the customer.
How we use it
When we start talking about RSAT with the customer, they always reply, that there is no budget. This is not correct, there are always costs on testing. Only the current test cost is hidden within the organization of that customer in the form of internal tests and/or issues arising in production environments. What we see is that the customer has a budget for his IT projects. The hours that the customer/key user is testing is not part of this budget.
Then we see another point, the current task recordings are most of the time outdated (see blog Why should your Business Process Modeler (BPM) be up to date?). By using RSAT you are automatically being forced to bring your BPM library up to date.
RSAT can become even more powerful by using it like a provisioning tool. Below is an example as we use it in a release pipeline. The first example is when we restore a database on a tier 1 box (MS hosted, or customer hosted)
The current RSAT solution needs to be installed on an actual machine instead of running as a cloud service. To achieve this we installed it on a cloud hosted Devbox and turned on auto shutdown to manage costs. Therefor you see steps both starting and stopping said environment in the train.
The current RSAT on a devops only runs on a real machine where it is installed. So, we installed it on cloud hosted Devbox, to keep the Azure cost low we start that VM on demand. In the Parameters of the RSAT provisioning we select the RSAT settings file pointing towards the environment we just restored the database of.
The tasks we use RSAT for are:
- Enable users
- Start batches
- Update exchange settings
- Any other integration options
But RSAT is also reusable when you deploy a new release with your latest tailored customer code. The reason I use RSAT this way is, testing on a Tier 1 is not representing the world, I want to test the Package in the Operational area and not in the Development area.
And the last case is for updates of Microsoft. It is similar like above , you only run the RSAT task on the MS release candidate environment.
So now that the business case has been explained, let us have a short look at the building blocks.
For starting and stopping the VM we use Azure CLI
Agent Pools
The steps we used need different Agent pools. I always prefer to use only the azure Pipelines. But those have a limitation. In case the step needs programs or files that are located on the VM, the VM also need to become an agent pool. Below list shows the additional agent pools we need at a customer implementation
And when you realize you need a building pool on an existing cloud hosted or even running on a local VM, you can still install it. Look for information on https://docs.microsoft.com/en-us/azure/devops/pipelines/agents/v2-windows?view=azure-devops
And now to wrap up; RSAT does not need a Sales story. If you know how to use it, It will start selling itself.
For other DEVOPS tips have a look at https://kaya-consulting.com/category/lcs/
Is your Business Process Modeler updated?
I see a lot of MS Dynamics 365 implementations where the BPM in Lifecycle Services (LCS) is not used. They are running fine (at least that is what the clients are telling me). But in general, their implementation partner did not tell them all the benefits LCS has in store for them. Why should a partner bother to mention all the benefits Improvements on testing and user help experience does not result in more billable hours? So is your Business Process Modeler updated?
Using the BPM in combination with DevOps supports more streamlined and efficient implementation projects. By defining the processes in the BPM library; there is a generic structure for tracking the project requirements / user stories / tasks. The BPM hierarchy can be synchronized with DevOps – as a result the DevOps backlog and boards are populated with structured work items ready to be used by the project team.
In addition to tracking the progress of a project, the BPM and DevOps processes and work items can be used for automated testing (RSAT) and user documentation (online help recordings). Read more on these topics in our other blogs!
When defining the processes and work items it is important to understand how LCS and DevOps share the information. There are various approaches one can take regarding the level of detail to track in BPM vs DevOps.
Business Process Modeler
The BPM is a necessary evil – I must admit, it is a strange thing. Currently it is the only way to bring business processes, task recordings, automatic testing, and user documentation of Dynamics 365 together and re-use the information between the different areas. At the same time, it is not extremely user friendly when compared to for example DevOps.
recommendations
- The model name must be short, the name is reused everywhere later-on in DevOps, I always call it simply BPM, so later the titles of my epics, features and tasks start with [BPM]
- The existing BPM models from MS are too big and complex, no key user will ever be able to put the task recording on the correct place.
- Create your own BPM
- 2 layers
- The 3 layer is in additional point for adding the task recording. A BPM could be like below picture, the customer has 2 tasks recording on how to create a customer
The BPM can then be synchronized with DevOps – the above hierarchy will generate DevOps work items with type “Epic” and/or “Feature” based on the synchronization parameters.
Once the epics and features have been synchronized, it is possible to add Requirements per process node via the button “Add requirement” – the work item type of a requirement in DevOps is also defined in the synchronization settings of LCS and depend on the DevOps project process template.
All links to DevOps are visible in LCM on the BPM nodes, including the test cases. The below considerations are important when making the decisions regarding what type of work items to track in which application.
BPM and DEVOPS
How does it integrate
- The synchronization between LCS and DevOps is one-directional from LCS BPM to DevOps, as a result:
- Any updates made in DevOps will not be reflected in the BPM library
- All work items that need to be available in the BPM need to be created via BPM and synchronized to DevOps
- Business processes (epics, features) need to be created in BPM, not in DevOps
- The mapping between BPM and DevOps is done in LCS and the options available depend on the project template type used for the DevOps project:
- when “Agile”, a BPM requirement type will generate a work item with type “User story” in DevOps
- when “Scrum”, a BPM requirement type will generate a work items with the type “Task” in DevOps
- If the DevOps template is “Custom” – including when it is derived from Agile – a BPM requirement type will generate a work items with the type “Task” in DevOps
- Test cases are linked to BPM process nodes (epic/feature in DevOps), not the requirements (user story/task in DevOps), maximum one test case can be linked to one process
- The one-directional synchronization rule applies also here
- If the test case should have a link to the requirement / user story / task, this can be added manually in DevOps
The DevOps process templates can be defined via DevOps Organization setup screen. The standard process templates are pre-created (Basic, Agile, Scrum, CMMI) and can be extended (KANBAN below).
The process template must be linked to the DevOps project via project settings screen:
The synchronization settings between LCS and DevOps are configured via LCS project settings tab “Visual Studio Team Services”. The options available in this screen depend on the process template used for the selected project.
In case of Agile process template, it is possible to select “User story” in the column “VSTS work item type” for LCS work item sub-type “Requirement”:
In case of Custom process template, it is not possible to select “User story” in the column “VSTS work item type” for LCS work item sub-type “Requirement”. The only available options are “Task” and “Bug”:
For additional details regarding the different DevOps process templates and how these are intended to be uses please refer to the Microsoft documentation
Conclusion
Considering the usability of LCS BPM and DevOps and the synchronization limitations our recommendation is to use BPM for 2-3 levels of processes and use DevOps for tracking the user stories / requirements / tasks. With this approach, the end users and key users can add new user stories / requirements via DevOps and link them to the proper BPM nodes (features / epics) inside DevOps.
When a new process is added, this must first be created via LCS BPM and synchronized to DevOps to make it available as a parent on new user stories / requirements.
Good luck on your improved user experience!
Automated backups of D365FO databases.
When you work with LCS then below list of data base movement operations are familiar to you. Nowadays exporting and refreshing can also be triggered with the Database Movement API This can become extremely handy for daily automated backups of D365FO databases. How we do it? With DEVOPS release pipelines. The pipeline looks like this. The BACKUP TST is disabled because the API does not ‘yet’ support Tier 1 Databases.
Speeding up
Because we had to apply automated backups of D365FO databases at multiple customers. It became time consuming. So, we optimized it. All customer specific setting became Variables of the pipeline. Because customer specific Client ID, Client Secret, and password, are sensitive information we have hidden them. Be aware in case you unhide, the values are cleared.
PowerShell
Below PowerShell script is a sample of how we generate the Token. This token is later used on the backup commands PowerShell.
| $tokenUrl = “https://login.microsoftonline.com/common/oauth2/token” $tokenBody = @{ grant_type = “password” client_id = “$(CLIENTID)” client_secret = “$(CLIENTSECRET)” resource = “https://lcsapi.lcs.dynamics.com” username = “$(USERNAME)” password = “$(PASSWORD)” } $tokenResponse = Invoke-RestMethod -Method ‘POST’ -Uri $tokenUrl -Body $tokenBody $token = $tokenResponse.access_token Write-Host “##vso[task.setvariable variable=TOKEN] $token” |
In the next sample we create the backup
| $cstzone = [System.TimeZoneInfo]:ConvertTimeBySystemTimeZoneId( (Get-Date), ‘W. Europe Standard Time’) $filedate = Get-Date $cstzone -f “yyy-MM-dd” $BackupName = “ACC-$filedate” Write-Output $BackupName $refreshUrl = “https://lcsapi.lcs.dynamics.com/databasemovement/v1/export/project/$(LCSPRPOJID)/environment/$(ACC)/backupName/$BackupName” $refreshHeader = @{ Authorization = “Bearer $(TOKEN)” “x-ms-version” = ‘2017-09-15’ “Content-Type” = “application/json” } $refreshResponse = Invoke-RestMethod $refreshUrl -Method ‘POST’ -Headers $refreshHeader Write-Output $refreshResponse |
And the DEVOPS insider will comment, yeah dream on, no build , is also no release. True, you just must reuse a build pool in your environment.
Please be aware that MS is throttling these requests, only 3 backups a day are possible.
And at the end, we generate an email notification for the people that really need it.
LCS
We do not use the alerts from LCS, it causes simply too many emails. So, we disable the next notification for all LCS users
Don’t forget the other handy LCS tricks
Performance tuning development VM
It is a general complain of any developer. Why is it so SLOOOWWWWW? And yes, it is slow. But there are performance tuning tricks. Slow is influence on 3 things
- Startup VM
- Is also startup of
- Batches
- Management reporter
- Windows security
- Windows update
- Windows maintenance
- SQL
- There is no SQL maintenance
- Disk
- Is also startup of
You could disable the batches & management reporter. That will help. But the real issue is how the disks of that VM are configured. In general, it looks like below 16 lazy disks.
Now you can change these to faster disks , but please be aware of the prices. Also, these are managed disk, the cost fixed amount every month, so stopping the VM will not reduce below cost.
| Size | Premium SSD | Standard SSD | Standard HDD |
| 64 GiB | 9.47 | 4.05 | 2.54 |
| 128 GiB | 18.29 | 8.10 | 4.97 |
| 256 GiB | 35.26 | 16.20 | 9.56 |
| 512 GiB | 67.92 | 32.39 | 18.36 |
How does the disk migration work? The first step is migration to managed disks
Managed Disks simplify disk management for VMs by managing the Storage accounts behind the scenes. Managed Disks also provide granular access control with RBAC and better reliability for VMs in an Availability Set. Learn more about the benefits of using Managed Disks
Source unmanaged disks are not deleted after the migration. Managed Disks are created by making a copy of the source disks. You can revert back to unmanaged disks by creating a new VM with the source disks. Configuration of the VMs is not changed after the migration. Learn more about migrating to Managed Disks
In order to complete migration, we will need to start the virtual machine. Once migration is complete, you may stop the virtual machine.
Next you can change all the discs
But hold on 16 disks multiplied with 18.29 makes 292 euro on top of running the VM. This is very expensive in the relation to how you use it. So next step:
- Archive your code & database.
- Delete the VM
- Deploy a new one
And of course, do not forget to setup the auto shutdown and the azure software reductions
Dynamics 365 LCS Tricks – install Knowledge Base (KB) article
LCS is the place for getting all the information from Microsoft related to your Dynamics 365 for Operations projects. This Web portal also allows you to get your MS fixes to your environment and knows which Knowledge Base (KB) articles have been applied.
In the next image, for instance, we can see that there are 167 fixes applicable. So, let’s start cherry picking.
Dynamics 365 LCS tricks – platform update 4 experience
Dynamics 365 LCS Tricks – mass install packages
Deploying packages is a time-consuming thing. MS has improved the speed, but we partner can also help. This helping is by merging the packages. Good merged combinations are Binary updates and x++ updates, but your own code can also be combined with your customizations on your Retail (SDK) part.
Till now I was not able to merge ISV packages, so those must be deployed separately. However, there is quick dirty trick for local VM.
Dynamics 365 LCS tricks – force LCS to update
LCS is your friend in Dynamics 365 as long it is up-to-date. And I have seen it is not always the case. So I will share some tricks to force it.
There are some things you should know LCS is only updating 1 time a day and that is 10 o clock in the evening. But when you look in LCS you see it is not updated for a couple of days. So what’s next?
Dynamics 365 LCS tricks – A day in the life of LCS
We all must use LCS when working with Dynamics 365 these days, sometimes it feels good and sometimes it is frustrating.
The next real live image of LCS shows how I feel about the experience.
We are speeding up but at the same time we have to wait.
