The time has come to start a new part of my worklife! I am leaving Basefarm and the Lead Architect role I had there for a new job as a Chief Cloud Architect at Evry.
During my time at Basefarm I have been working intensively with Azure Stack and Azure offerings as well as on-premise solutions. Basefarm was one of the Azure Stack early adopters and the learnings within this journey have been challenging and interesting both based on the appliance and also the organisational adoption.
My key areas as a Chief Cloud Architect will be on public cloud solutions and helping my team at Evry and the customers be successful in the transformation to the public cloud.
I will be focusing on amazing solutions based on Azure and Azure Stack but I will also work on AWS and Google Cloud solutions.
I will have the great pleasure to work with Marius Sandbu that is also a Microsoft Most Valuable Professional.
We have an upgraded Hyper-V cluster from 2012 R2 to 2016 and I wanted to make some more space on the volume that it resides on so I started looking on the settings of the VMs. As I concluded in another blog post there can be a discussion about if we really need the setting of “Save VM state” for the VM´s in a Hyper-V cluster. It is viable to have this setting on VM´s that reside on a standalone host and if doing maintenance it saves the state of the VM during host reboot.
So what is the thing here, well some of the VM´s in this cluster have got their VM version upgraded to 8.0 but some was still on 5.0. If i just check on the SC VMM I could find the VM´s configured with “Save VM” and amount of storage that they consume.
But when checking the storage where the VM reside I noticed that the value above did not match the size. With the following PowerShell I check for the files that have the ending of .bin (2012 r2 and older format) and also .vmrs (that is the 2016+ format)
Running and operating an Azure Stack either on a DevKit or a integrated system can be a hurdle and sometimes you need to know the state of the stamp and the portal does not always show everything.
Connecting an session to a emergency recovery console and kicking of a Test-AzureStack can give you some more insights to what is the state of the system.
If you want to know more and see the state you better look at the parameters of Test-AzureStack because there are some hidden gems there! If you run with a -ServiceAdminCredential firstname.lastname@example.org you will get some information and see what actually works on the stamp in regards of deployment and usage of the base RP´s.
If you do not want to run all Test-AzureStack tests you can specify running just -Include AzsScenarios and thus only running Operator and User Scenarios and not all other tests with fabric and storage etc. There is another parameter -timeout that can be used if you need more time for the test to run
One thing to consider is that it is cumbersome to utilize a serviceadmin credential that is MFA-enabled for the Test-AzureStack and that you have to set up a separate account for this test.
Now finally Microsoft have updated the misleading documentation on supported guest os within Hyper-V. This is quite important as some people tend to get stuck on small details and as my good friend Didier wrote on his blog, Hyper-V supports guest OS n+1, although that now gets a bit altered with the semi-anual releases.
Now the docs page is updated and shows the following:
Automatic VM activation heads-up!
There is though a small or big thing that needs to be considered if you have an environment with Hyper-V servers and utilize the AVMA. If you plan to deploy Server 2019 guest VM´s there is no way to get them auto activated on a 2016 Hyper-V host.
If you are a bit more old fashioned and utilize a KMS you will just need a KMS server that is newer than 2012 as the key for 2019 needs the KMS to be hosted on at least a 2012R2 Windows Server!
My homelab environment consists of two Intel NUC and I have been playing around with the insiders previews of Server 2019 on one of them and the other one was running Server 2016.
As you might know there is a bit of a hustle of the nic drivers with the server versions of Windows and the Intel NUC´s so there are some steps to get it working. I had some issues where the nic failed during in-place upgrade between preview versions of 2019 and as I do not have a KVM I had to move the NUC to a monitor and fix it. To get the drivers in I had to set the server into test mode:
bcdedit/set LOADOPTIONS DISABLE_INTEGRITY_CHECKS
bcdedit/set TESTSIGNING ON
bcdedit/set NOINTEGRITYCHECKS ON
After I did this and rebooted the server I could update the nic drivers that I already had modified as per this blog post.
I wanted to test and update my 2016 server with an in-place upgrade without moving it from the closet and as a precaution I changed to test mode first and then started the update…
After the upgrade went through successfully I changed back to non-test-mode:
bcdedit/set LOADOPTIONS ENABLE_INTEGRITY_CHECKS
bcdedit/set TESTSIGNING OFF
bcdedit/set NOINTEGRITYCHECKS OFF
I had a small issue with the Windows Update after the upgrade and it would not finish installing the CU of 2018-12… As a mitigation if this I went for the command line tool of System File Checker, SFC and the parameter /Scannow I also did a Dism repair and after these two successfully ran I could continue with the Windows Update!
The Azure Stack have several Resource Providers that can be utilized to bring value to the stack.
We have an Azure Stack in the company and have had an early adopter experience. To make the most of our testing and offering we added App Service, MySQL, SQL RP´s after deployment.
In our multi-tenant usage registered Stack we noticed on our own bill that it was a bit high $$ and realized that it was the shared workers that was behind this, during a couple of Azure Stack work shops we had scaled it to 12 instances for labs. The shared workers are billed to the registration CSP-subscription, the dedicated are billed to the customers subscriptions when they are in use in an app plan but you as a stack provider can have several of them running and prepared without any extra cost.
If you want to add or remove worker instances this can be mitigated with PowerShell or through the portal:
There are some caviats about this and that can be read in the app service documentation, if you want to give the user subscriptions access to serverless functions on a consumption plan you have to have enough shared workers available…. read more here
Carefully monitor the capacity and usage of your add-on RP´s so the experience for the customer always is great!
Last Friday Bruce Payette from Microsoft published the prerelease of 1.0.0 of Import-WindowsCompatibility module that you can use for being able to utilize Windows PowerShell modules in PowerShell Core
To find and install the prerelease from the PowerShell gallery you have to add a parameter in the install-module command and that is -allowprerelease.