Monday, January 26, 2015

Reconfigure your Resource Providers for Azure Pack

Configuring your Resource Providers for Azure Pack

While deploying Windows Azure Pack, several factors plays its part when it comes to the design and layout of the solution. As you may be aware of, Windows Azure Pack contains a lot of different sites, APIs and resource providers – just so that you can enable and realize Azure technologies within your own datacenter.
It’s more than a glorified self-service portal so the requirements for the design, load and can be overwhelming for some customers.

Before I get to the big point of this blog post, I would like to put it into some context first.

Normally at customer sites, we see the following different designs when it comes to Windows Azure Pack.

Express



Organizations who want to just test and play around are deploying the single install, express setup of Windows Azure Pack. This will install all the sites and APIs onto a single virtual machine, and the organizations can easily add resource providers to start testing the powerful cloud enabled tool.

Although I have seen some examples where the Express setup has been used in production, it is far from what we recommend. The public facing parts of Azure Pack, such as the Tenant Public API, Tenant Site and eventually Tenant Authentication Site are directly exposed on the internet. Having everything on the same virtual machine will increase the attack surface as well as lead to performance, HA and scale issues.

Configuration requirements using this design:

There aren’t any hard requirements using the Express solution as we like to think that people are only using it in lab and test. However, if you want to make it available and actually use it across firewalls, you will have to perform the following:

·       Reconfigure tenant site (FQDN, certificate and port)
·       Reconfigure tenant authentication site (FQDN, certificate and port)
·       Reconfigure tenant public API (FQDN, certificate and port)

Optional:

·       Reconfigure admin site (FQDN, certificate and port)
·       Reconfigure admin authentication site (FQDN, certificate and port)


Basic



For some of the smaller customers where HA is not the most important thing, we often see a basic implementation of Windows Azure Pack. This means that we have a single virtual machine running the high-privileged services – such as the Admin API, Admin Site, Tenant API and eventually Admin Authentication site together with the default resource providers. This virtual machine is located behind the firewall and in most cases within the same Active Directory Domain with its resource providers (SCVMM+SPF, SQL, ServiceBus, WebSites etc).

For the public facing part (the parts mentioned before, directly exposed on the internet) they use another – dedicated virtual machine which might be located in DMZ and available on the internet.
Of course, both the high-privileged VM and the internet facing VM are running on a Hyper-V cluster so that the VMs themselves are highly available.

Configuration requirements using this design:

I strongly recommend using a highly available WAP design whenever you plan to put it into production. But in this design, the only presence of HA is at the hypervisor level.
You will have to perform the following using this design:

·       Reconfigure tenant site (FQDN, certificate and port)
·       Reconfigure tenant public API (FQDN, certificate and port)
·       Reconfigure tenant authentication site (FQDN, certificate and port)
o   Or
·       Integrate with Active Directory Federation Services and remove tenant authentication site

Optional:

·       Reconfigure admin site (FQDN, certificate and port)
·       Reconfigure admin authentication site (FQDN, certificate and port)
o   Or
·       Integrate with Active Directory Federation Services and remove admin authentication site


Minimal Distribution



The most common design of Windows Azure Pack and what’s normally at least what I am recommending, Is where we have at least two virtual machines for the high-privileged servers, configured as highly available behind a load balancer, and the same for the internet facing part.
This will indeed require load balancers and VIPs, but also some additional reconfiguration when it comes to the Azure Pack environment.

Configuration requirements using this desing

Having the high-privileged services as well as the internet facing parts scaled across several virtual machines, helps us to address performance, availability and scale issues.
You will have to perform the following reconfiguration to make this work:

·       Reconfigure tenat site (FQDN, certificate and port)
·       Reconfigure tenant public API (FQDN, certificate and port)
·       Reconfigure tenant authentication site (FQDN, certificate and port)
o   Or
·       Integrate with Active Directory Federation Services and remove tenant authentication site
·       Reconfigure admin site (FQDN, certificate and port)
·       Reconfigure admin authentication site (FQDN, certificate and port)
o   Or
·       Integrate with Active Directory Federation Services and remove admin authentication site
·       Reconfigure resource providers (FQDN and certificate):
o   Sqlserver
o   MySQL
o   Monitoring
o   Marketplace
o   Usageservice
o   Systemcenter
o   Webspaces
o   Servicebus


Summary

So whenever you plan to scale out and ensure HA across all sites and APIs, you have to reconfigure the components as mentioned with the Minimal Distribution design. The same rules apply if you intend to be more drastic around this, having dedicated VMs for each and every site and API. The reconfiguration is still mandatory.

Windows Azure Pack has been available for over a year now, and the majority of organizations are adopting the VM Cloud resource provider. The good thing here is that even if you have scaled out the SPF endpoint, you are simply adding the endpoint to the admin API and everything is handled.
There’s really not much reconfiguration required if you have configured SPF correctly with FQDN and certificates upfront.

What’s more of a concern is when you want to add resource providers such as SQL server(s) and/or MySQL server(s).

By default, when you install the first high-privileged server with the admin API, admin site and so on, you also get the default resource providers added, such as SQL, MySQL, Usage, Monitoring, Servicebus and Marketplace. The FQDN’s are bound to the computer name of this machine.
Once you add the second – or even third VM that should be located behind a load balancer together with the first VM, these resource providers must also be reconfigured so that you are not pointing toward an individual virtual machine, but towards a FQDN that is associated with a VIP behind the load balancer.

Reconfiguring the default Resource Providers – and why that can be a pain

In order to reconfigure the Windows Azure Pack portals, APIs and resource providers, we have to instrument the databases in a supported way. The supported way is through Powershell, and together with my good friend Flemming Riis, we have convered how to reconfigure the high-privileged services – as well as the internet facing parts in some earlier blog posts.

As a result of that, I won’t cover it over again, but rather refer to those URL’s, hoping you will notice them, read them and continue reading this blog post as I am about to reach my point.





Allright, let us continue on the resource providers.

You are probably familiar with the reconfiguration of the tenant and admin stuff by now, and understand that we have several sets of APIs and portals involved. In the end of the day, everything here should interact nicely together, being able to reach each other and expose the right set of information to both an administrator and a tenant.

If we look at the resource providers we are dealing with directly in the database, we can see that we have several endpoints to each and every resource provider.
We have an endpoint for the resource provider when it comes from the admin API, and we have an endpoint for the resource provider when coming from the tenant site and API.
In addition, each resource provider have an endpoint for usage and notification too.



Example:

The SQLserver resource provider will have the following endpoints:

AdminEndPoint.ForwardingAddress: https://FQDN:30010/
TenantEndPoint.ForwardingAddress: https://FQDN:30010/Subscriptions/
UsageEndPoint.ForwardingAddress: https://FQDN:30010/
NotificationEndPoint.ForwardingAddress: https://FQDN:30010/

So, when only reconfiguring the FQDN, certificates and ports for the high-privileged services and internet facing parts, these endpoints are left behind.

The same is applicable for the other resource providers as well, an in order to turn the in to highly available resource providers, you must perform this through powershell:

Example:

Since ADSF is added to the mix, we need to create a function that will allow us to get the token we require for accessing the AdminURI.

function Get-AdfsToken([string]$adfsAddress, [PSCredential]$credential)
{
    $clientRealm = 'http://azureservices/AdminSite'
    $allowSelfSignCertificates = $true

    Add-Type -AssemblyName 'System.ServiceModel, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089'
    Add-Type -AssemblyName 'System.IdentityModel, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089'

    $identityProviderEndpoint = New-Object -TypeName System.ServiceModel.EndpointAddress -ArgumentList ($adfsAddress + '/adfs/services/trust/13/usernamemixed')
    $identityProviderBinding = New-Object -TypeName System.ServiceModel.WS2007HttpBinding -ArgumentList ([System.ServiceModel.SecurityMode]::TransportWithMessageCredential)
    $identityProviderBinding.Security.Message.EstablishSecurityContext = $false
    $identityProviderBinding.Security.Message.ClientCredentialType = 'UserName'
    $identityProviderBinding.Security.Transport.ClientCredentialType = 'None'

    $trustChannelFactory = New-Object -TypeName System.ServiceModel.Security.WSTrustChannelFactory -ArgumentList $identityProviderBinding, $identityProviderEndpoint
    $trustChannelFactory.TrustVersion = [System.ServiceModel.Security.TrustVersion]::WSTrust13

    if ($allowSelfSignCertificates)
    {
        $certificateAuthentication = New-Object -TypeName System.ServiceModel.Security.X509ServiceCertificateAuthentication
        $certificateAuthentication.CertificateValidationMode = 'None'
        $trustChannelFactory.Credentials.ServiceCertificate.SslCertificateAuthentication = $certificateAuthentication
    }

    $ptr = [System.Runtime.InteropServices.Marshal]::SecureStringToCoTaskMemUnicode($credential.Password)
    $password = [System.Runtime.InteropServices.Marshal]::PtrToStringUni($ptr)
    [System.Runtime.InteropServices.Marshal]::ZeroFreeCoTaskMemUnicode($ptr)

    $trustChannelFactory.Credentials.SupportInteractive = $false
    $trustChannelFactory.Credentials.UserName.UserName = $credential.UserName
    $trustChannelFactory.Credentials.UserName.Password = $password #$credential.Password

    $rst = New-Object -TypeName System.IdentityModel.Protocols.WSTrust.RequestSecurityToken -ArgumentList ([System.IdentityModel.Protocols.WSTrust.RequestTypes]::Issue)
    $rst.AppliesTo = New-Object -TypeName System.IdentityModel.Protocols.WSTrust.EndpointReference -ArgumentList $clientRealm
    $rst.TokenType = 'urn:ietf:params:oauth:token-type:jwt'
    $rst.KeyType = [System.IdentityModel.Protocols.WSTrust.KeyTypes]::Bearer

    $rstr = New-Object -TypeName System.IdentityModel.Protocols.WSTrust.RequestSecurityTokenResponse

    $channel = $trustChannelFactory.CreateChannel()
    $token = $channel.Issue($rst, [ref] $rstr)

    $tokenString = ([System.IdentityModel.Tokens.GenericXmlSecurityToken]$token).TokenXml.InnerText;
    $result = [System.Text.Encoding]::UTF8.GetString([System.Convert]::FromBase64String($tokenString))
    return $result

Next, we will list the resource providers we have and query for the endpoints:


### Change the variables to fit your environment

$adfsAddress = 'https://adfsserver.com
$username = 'domain\username'
$password = 'P@@$Word'
$adminuri = ‘https://adminapi.domain.com:30004’
$securePassword = ConvertTo-SecureString -String $password -AsPlainText -Force
$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList $username,$securePassword

$token = Get-AdfsToken -adfsAddress $adfsAddress -credential $credential

$token

$FQDN = ‘https://wap.domain.com’

### Get a list of all your resource providers

Get-MgmtSvcResourceProvider -IncludeSystemResourceProviders -AdminUri $adminUri -Token $token -DisableCertificateValidation | Format-List -Property "Name"

Get-MgmtSvcResourceProvider -IncludeSystemResourceProviders -AdminUri $adminuri -Token $token -DisableCertificateValidation | fl -Property "name"

# Get a list of resource providers with the current configured endpoint values
$rp = Get-MgmtSvcResourceProvider -IncludeSystemResourceProviders -AdminUri $adminUri -Token $token -DisableCertificateValidation
$rp | Select Name, @{e={$_.AdminEndPoint.ForwardingAddress}}, @{e={$_.TenantEndpoint.ForwardingAddress}}, @{e={$_.UsageEndpoint.ForwardingAddress}}, @{e={$_.healthcheckendpoint.forwardingaddress}}, @{e={$_.notificationendpoint.forwardingaddress}}

# STEP 1 - Configure new FQDN for the SQLserver resource provider

$resourceProviderName = "sqlservers"
$adminEndpoint = "https://$FQDN:30010/"
$tenantEndpoint = "https://$FQDN:30010/subscriptions"
$usageEndpoint = "https://$FQDN:30010/"
$healthCheckEndpoint = $null
$notificationEndpoint = "https://$FQDN:30010/"


$rp = Get-MgmtSvcResourceProvider -Name $resourceProviderName -IncludeSystemResourceProviders -AdminUri $adminUri -Token $token -DisableCertificateValidation
# update all the endpoints using the new fqdn:
if ($rp.AdminEndpoint -and $adminEndpoint) {
# update endpoint
$rp.AdminEndpoint.ForwardingAddress = New-Object System.Uri($adminEndpoint)
}
if ($rp.TenantEndpoint -and $tenantEndpoint) {
# update endpoint
$rp.TenantEndpoint.ForwardingAddress = New-Object System.Uri($tenantEndpoint)
}
if ($rp.UsageEndpoint -and $usageEndpoint) {
# update endpoint
$rp.UsageEndpoint.ForwardingAddress = New-Object System.Uri($usageEndpoint)
}
if ($rp.HealthCheckEndpoint -and $healthCheckEndpoint) {
# update endpoint
$rp.HealthCheckEndpoint.ForwardingAddress = New-Object System.Uri($healthCheckEndpoint)
}
if ($rp.NotificationEndpoint -and $notificationEndpoint) {
# update endpoint
$rp.NotificationEndpoint.ForwardingAddress = New-Object System.Uri($notificationEndpoint)
}

Set-MgmtSvcResourceProvider -ResourceProvider $rp -AdminUri $adminUri -Token $token -DisableCertificateValidation -Force

# Repeat STEP 1 and STEP 2 on the remaining resource providers



 By following the steps in this blog post, you shouldn’t have any warnings or errors in your WAP portals.

No comments: