Azure Insights: AKS bursts; AD Connect PHS issues; Hyper-V vault recovery error

Azure pros discuss AKS bursting when a node pool is full, AD Connect PHS issues, and sorting out vault recovery errors with Hyper-V.

Managing AKS bursts when a node pool is full

Microsoft MVP Richard Hooper, writing on Pixel Robots, described a recent question he received through LinkedIn asking if it is possible to do a Kubernetes deployment on an AKS node pool when the node pool is full burst for Container Instances. Using preferredDuringScheduling

IgnoredDuringExecution, Hooper experimented on AKS. He shared a basic YAML deployment manifest with some additional code added to it.

A special affinity section defines the agent pool and label type, while additional lines indicate that pods in the deployment can be scheduled with a node of a particular taint. The process works well for scaling up but can be challenging to scale down. He wrote:

If you have a system node pool, user node pool, virtual node the preferred scheduling order for this deployment will be the user node pool defined in the node affinity and then the virtual node. If you are using the system node pool and virtual node only then the system node pool will be preferred. If you have a system node pool, user node pool and a virtual node and are trying to schedule the pods on the system node pool then once the pods are running on a virtual node and you want to scale down, they will scale down from the system node pool first. This is due to the system node pool being important to the running of the AKS cluster.

Solving PHS issues with AD Connect

Microsoft MVP Sander Berkouwer, writing on The Things That Are Better Left Unspoken, discussed the new AD Connect release. The new version was updated to resolve a security issue, as well as problems with Password Hash Synchronization transactions. Previously, organizations using AD Connect sometimes noticed PHS stop working, due to a faulty library. He stated: