microsoft / cluster Goto Github PK

To avoid accidental deletion of parent nodes, [ClusterResourceGroup]::Create() should lock a resource group when it is created. Cluster should also implement [ClusterResourceGroup]::Delete() that deletes a resource group and all its descendants in parallel, abstracting away all locks.

Either some pre-deployment validation is required in Cluster::PublishConfiguration or Publish-ClusterConfiguration should remove the default value for DefinitionsContainer parameter

Cluster requires "VhdContainer" and "ImageUrl" template parameters to deploy because we can't delete the associated resources.

Azure Traffic Manager integration should be supported out of the box

It's possible for a deployment to succeed in the console, then fail validation immediately in the Azure RM system.

Repro: Choose a VM size not available in a region and attempt to deploy it. As internal users, this is any VM in West US 2.

Perhaps some prevalidation can occur?

Delete all DSC events from the event log before image baking to avoid confusion

This is potentially an AzureRM bug, but DSC deployments have been failing and errors are only listed in portal instead of in the terminal

Right now, storage accounts and key vault names are completely random at all levels. It would be nice if it would reflect at what level of the topology hierarchy it exists at, and make it much easier to find in azure portal.

Clusters should reference resources in their parent Environment to avoid copy time when creating a new Cluster.

$scripts: is broken. As a workaround, I populated this field, which made things happy.

ClusterResourceGroup::PropagateBlobs(...) should use ClusterResourceGroup::GetDescendants() to propagate all blobs in parallel instead of iterating across each child after the parallel copy

Much of the code is uncommented and it is not immediately apparent what it's function is.
Example:

 [Cluster] NewChildCluster() {
        $indexes = ($this.GetChildren() | % {[Cluster]::new($_)}).Index
        for ($index = 0; $index -in $indexes; $index++) {}
        $cluster = [Cluster]::new("$this-$index")
        $cluster.Create()
        return $cluster
    }

New-AzureRmResoureGroupDeployment throws multiple errors in an ArrayList that serialize to Conflict, losing the underlying messages of each error. No iterables should be thrown--they should all be unwrapped into an AggregateException

Multiple error occurred: Conflict,Conflict,Conflict,Conflict,Conflict. Please see details.
At C:\Program Files\WindowsPowerShell\Modules\CLuster\0.1.3\Types.ps1:483 char:13
+             throw $deploymentErrors
+             ~~~~~~~~~~~~~~~~~~~~~~~
    + CategoryInfo          : OperationStopped: (System.Collections.ArrayList:ArrayList) [], RuntimeException
    + FullyQualifiedErrorId : Multiple error occurred: Conflict,Conflict,Conflict,Conflict,Conflict. Please see details.

Proposed function:

Test-ClusterTopology [-ClusterSet] <ClusterResourceGroup>

Determines whether all required resource groups under/including a ClusterSet exist with the correct resources.

This should catch all cases where a resource group was created but not correctly populated, as well as cases where a parent resource group was deleted while a child exists.

I had thought we were sorting by modified date, but that does not appear to be working as intended.

Right now, only the user who created a key vault can manage it