How to revert new Transparent Page Sharing behaviour

There have been a lot of blog posts around warning us that starting from vSphere 5.5 Update2d (build 2403361) TPS will be basically turned off.

Besides comments about the why and how to partially re-enable inter-VM TPS I just think in most cases it might make sense to restore the original behaviour; I have been in the process of upgrading an infrastructure where the RAM contrain represented a much bigger concearn than the specific security reason why TPS has been turned off so I decided to take a look at the shared memory metrics in vCenter just to find out that post-upgrade situation wouldn’t be exactly heavenly if all the shared memory pages would be translated into unique pages.

I found this arcicle KB2097593 where it explains the various status of the settings you can apply and one of them basically disables the new behaviour restoring the old fashion TPS:

Featured image

In other words setting the value “ShareForceSalting” to 0 will bring things back in time.

The complete procedure is as follows:

1. Log in to ESX (i)/vCenter with the VI-Client.
2. Select ESX (i) relevant host.
3. In the Configuration tab, click Advanced Settings (link) under the software section.
4. In the Advanced Settings window, click Mem.
5. Search for Mem.ShareForceSalting and set the value to 0.
6. Click OK.
7. Reboot host

vSphere 6 Certificate Lifecycle Management

Recently I’ve been fighting with a vSphere environment and CA certificates and I thought a lot about certificate management and lifecycle in a VMware vSphere environment after that and how much it needs improvement. With the SSL Certificate Automation Tool VMware made a step in the right direction and even if the tools itself is sometimes a little buggy it is still very handy in automating a long and error prone process. In vSphere 6 VMware is taking another step in the right direction to help us create, apply and manage SSL certificates in a vSphere environment, but before talking about this we need to talk a bit about what’s new in SSO and vCenter architecture in vSphere 6. Since the introduction of SSO VMware changed its architecture in every major release, starting from 5.1 to 5.5 and now to 6.0 so let’s make a little bit of history:

Featured image

The new vSphere 6 management architecture introduces two main roles that you can deploy, these are the Management Node and the Platform Service Controller (PSC); the reason behind this separation is to have a logical entity that will take care of the main management features while another entity will hold the core and security features of the solution. What is nice about this separation is that you don’t need a 1:1 ratio between Management Nodes and PSCs so you can install PSC on separate boxes and replicate between them and then have as many Management nodes as you need (as long as you are within the same SSO domain)

Featured image

For HA scenario if you install PSC on separate boxes you will still need a load balancer. Supported solutions are Big-IP F5 and NetScaler so far.

You can obviously still install everything in one box:

Featured image

You might have noticed that the HA model for SSO was active/passive in 5.1, then active/active in 5.5 and now is active/passive again; this is due to the re-engineering of the Secure Token Service (STS) which is moving to a new and more robust method of STS (known as WebSSO) which is the same already used by vCAC (or vRealize Automation if you will) and that will be used from now forward instead of the old 5.5 method (WS-Trust). Let’s see how services are spread out on each role:

Let’s take a look to the services within the Management Node and the PSC:

Featured image

In the Management Node we can find services and features that every vSphere Admin feels very comfortable and familiar with such as vCenter Server, vSphere Web Client, Syslog Collector, etc., but two of them deserve a few words:

  • Virtual Datacenter Service: this service is new and it has been introduced to help mitigate the limitation connected with the Datacenter object in vCenter as a Management boundary.
  • (Optional) vPostgres: This component is obviously referring to the vCenter Appliance (thus optional) but I believe more and more new deployments or upgrades deserve to be considered a good fit for vCSA since VMware announced complete equality of features between vCenter installed on Windows and vCSA; leave alone the fuss of dedicating Windows licenses for vCenter which might not be a huge problem I just find the process of patching ad upgrading a vCSA simply amazing and it’s not a secret that products like EVO:RAIL make extensive use of vCSA. VMware wants to move all their services deployment model towards Virtual Appliances, this is not a secret and we need to get used to it, the sooner the better, but I’m digressing…

Featured image

In the Platform Service Controller or PSC we find our old friend SSO (we have had a rough past but now we are on better terms) and quite a few new services:

  • VMware Single Sign-On
    • Secure Token Service (STS)
    • Identity Management Service (IdM)
    • Directory Service (VMDir)
  • VMware Certificate Authority (VMCA)
  • VMware Endpoint Certificate Store (VECS)
  • VMware Licensing Service
  • Authentication Framework Daemon (AFD)
  • Component Manager Service (CM)
  • HTTP Reverse Proxy

Describing all these services is out of the scope of this post but as you probably guess two of them will be our focus: the VMware Certificate Authority (or VMCA) and the VMware Endpoint Certificate Store (or VECS). But what are the roles of VMCA and VECS? The VMCA is no more or less than a CA, so you can:

  • Generate Certificates
  • Generate CRLs
  • Use the UI
  • Use the Command Line Interface to replace certificates

The VECS is where all certificates within the PSC are stored, with the only exception of the ESXi certificates that are stored locally on vSphere hosts, so here you can:

  • Store certificates and keys
  • Sync trusted certificates
  • Sync CRLs
  • Use the UI
  • Use the CLI to perform various actions

Since VMCA and VECS are part of the PSC, they will take advantage of the Multi-Master Replication Model which is offered by the Directory Service (VMDir) in order to achieve HA. In the past every service had its own user and required its own certificate but this is not the case anymore since we now have Solution Users (SU); since the number of services has increased significantly it would be impractical to manage the lifecycle of this many certificates so now we have 4 main SU that will hold the certificate used for a number of services.

Voila_Capture 2015-01-08_08-14-37_pm_white_background

What about use cases/scenarios in which I can implement VMCA? In what ways you can use this new tool?

Featured image

Scenario 1 and 2 are similar: the VMCA is the CA that releases certificates for all Solution Users (SU), the only difference is that in scenario 1 the VMCA is the root CA and you will need to distribute the Root CA Certificate so that all corporate browsers will trust it, while in scenario 2 the VMCA becomes part of an existing PKI as a subordinate CA and you certificate trust.

Featured image

In scenario 3 VMCA is installed but not used, CSRs are created and submitted to an external CA and VECS will be used to store certificates in PEM format.

Featured image

My favorite is scenario 2 because most enterprises I see already have a PKI (Microsoft CA usually) and all clients already trust the CA certificates, so adding the VMCA as s subordinate is a non disruptive process with a very low maintenance impact on the PKI itself, it protects investments already made to implement the current PKI and  preserves the knowledge to run the PKI.

Replacing certificates is still a CLI task (looks like Powershell will be involved) but VMCA and VECS are a very promising step toward the right direction for simplifying certificate lifecycle management in a vSphere environment.

Powershell Script for Shutting Down your vSphere Environment

Every now and then I need to setup an environment so that if a power outage occurs out of business hours there is a sort of automation taking care of that gracefully shutting down all VMs and Hosts to prevent failures.

For some reason I can never use older scripts I already have because whether they are too old and need some rewriting or because they don’t take into account some aspect of the specific environment I’m working on; one way or another it always feels like starting over.

While I was about to write a new one I stumbled across this blog post by Mike Preston and I was happy to find a very straightforward, simple and yet very complete script which would cover most of your (and mine too) needs.

Here are the aspects I was positively impressed about:

– it keeps into account HA
– it keeps into account DRS and stops it from moving things around while shutting down everything (nice)
– it allows you to specify if vCenter is virtual and it takes care of it for last, including the host where vCenter resides (nice!!)
– it keeps into account that VMs might just not shutdown gracefully for some reason; this is a major reason why most scripts would fail (see VDI streamed VMs)
– it writes events in a log file which is nice to go and look at in case you want to know what happened once you’re back online
– it dumps to file the list of powered on VMs at the moment of shutdown so you can manually turn everything back on as it was or take advantage of another nice script wrote by Mike called “poweronvms”

Ever if this script is definitely one of the most complete it is still lacking some of the requirements I needed, so with Mike permission (thanks Mike!) I added some things:

– added check for loading VMware PowerCLI Snap-In
– support for vApps
– removed annoying error messages when variables are = null
– added support for encrypted password dumped on a file (I don’t like password in clear text in scripts)
– added date-time in the log file for each event
– minor cosmetic in the output
– tested with vsphere 5.5 U1

Before we start using the script we need to create a password file so that passwords are not stored in clear text in the script itself:

mkdir c:\shutdown
Read-Host -Prompt "Enter password" -AsSecureString | ConvertFrom-SecureString | out-file c:\shutdown\cred.txt
(type in the password of the user you want to use for running the script)
cat c:\shutdown\cred.txt

The output should be something like this:

01000000d08c9ddf0115d1118c7a00c04fc297eb0100000092295625f5c35b4bb8af99be46e6679d0000000002000000000003660000c0000000100
000007b8d21c33686ef6ec751c0f671e7ff510000000004800000a00000001000000022d4534785ecda5799047d48f0f79a4618000000a1376ef3bf
fa5a3a928bee66d68eab5d49b12308f9d0b63f140000002ca923ea8eb54ff11c5864d9f722931a4ec85597

This is a hash of the password calculated using the current Windows credentials and it’s also connected with the Windows machine you run the command from, so this password file is not portable and can be used only but the user who generated it.

The script is already set up to use such password file, just edit the variables in the top of the script accordingly.

Here it is:

#################################################################################
# Power off VMs (poweroffvms.ps1)
#
# This script does have a 'partner' script that powers the VMs back on, you can
# grab that script at http://blog.mwpreston.net/shares/
#
# Created By: Mike Preston, 2012 - With a whole lot of help from Eric Wright
#                                  (@discoposse)
#
# Variables:  $vcenter - The IP/DNS of your vCenter Server
#             $username/password - credentials for your vCenter Server
#             $filename - path to csv file to store powered on vms
#			  used for the poweronvms.ps1 script.
#             $cluster - Name of specific cluster to target within vCenter
#             $datacenter - Name of specific datacenter to target within vCenter
#
#################################################################################

if ( (Get-PSSnapin -Name VMware.VimAutomation.Core -ErrorAction SilentlyContinue) -eq $null )
{
	Add-PsSnapin VMware.VimAutomation.Core
}

# Some variables
$vcenter = "<vcenter_fqdn>"
$username = "<vcenter_username>"
$password = get-content C:\shutdown\cred.txt | convertto-securestring
$credentials = new-object System.Management.Automation.PSCredential $username, $password
$cluster = "<cluster_name>"
$datacenter = "<datacenter_name>"
$filename = "c:\shutdown\poweredonvms.csv"
$time = ( get-date ).ToString('HH-mm-ss')
$date = ( get-date ).ToString('dd-MM-yyyy')
$logfile = New-Item -type file "C:\shutdown\ShutdownLog-$date-$time.txt" -Force

# For use with a virtualized vCenter - bring it down last 🙂 - if you don't need this functionality
# simply leave the variable set to NA and the script will work as it always had
$vCenterVMName = "NA"

Write-Host ""
Write-Host "Shutdown command has been sent to the vCenter Server." -Foregroundcolor yellow
Write-Host "This script will shutdown all of the VMs and hosts located in $datacenter." -Foregroundcolor yellow
Write-Host "Upon completion, this server ($vcenter) will also be shutdown gracefully." -Foregroundcolor yellow
Write-Host ""
Sleep 5

Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) PowerOff Script Engaged"
Add-Content $logfile ""

# Connect to vCenter
Write-Host "Connecting to vCenter - $vcenter.... " -nonewline
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Connecting to vCenter - $vcenter"
$success = Connect-VIServer $vcenter -Credential $credentials -WarningAction:SilentlyContinue
if ($success) { Write-Host "Connected!" -Foregroundcolor Green }
else
{
    Write-Host "Something is wrong, Aborting script" -Foregroundcolor Red
    Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Something is wrong, Aborting script"
    exit
}
Write-Host ""
Add-Content $logfile  ""

# Turn Off vApps
Write-Host "Stopping VApps...." -Foregroundcolor Green
Write-Host ""
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Stopping VApps..."
Add-Content $logfile ""

$vapps = Get-VApp | Where { $_.Status -eq "Started" }

if ($vapps -ne $null)
{
	ForEach ($vapp in $vapps)
	{
			Write-Host "Processing $vapp.... " -ForegroundColor Green
			Write-Host ""
			Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Stopping $vapp."
			Add-Content $logfile ""
			Stop-VApp -VApp $vapp -Confirm:$false | out-null
			Write-Host "$vapp stopped." -Foregroundcolor Green
			Write-Host ""
	}
}

Write-Host "VApps stopped." -Foregroundcolor Green
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) VApps stopped."
Add-Content $logfile ""

# Get a list of all powered on VMs - used for powering back on....
Get-VM -Location $cluster | where-object {$_.PowerState -eq "PoweredOn" } | Select Name | Export-CSV $filename

# Change DRS Automation level to partially automated...
Write-Host "Changing cluster DRS Automation Level to Partially Automated" -Foregroundcolor green
Get-Cluster $cluster | Set-Cluster -DrsAutomation PartiallyAutomated -confirm:$false 

# Change the HA Level
Write-Host ""
Write-Host "Disabling HA on the cluster..." -Foregroundcolor green
Write-Host ""
Add-Content $logfile "Disabling HA on the cluster..."
Add-Content $logfile ""
Get-Cluster $cluster | Set-Cluster -HAEnabled:$false -confirm:$false 

# Get VMs again (we will do this again instead of parsing the file in case a VM was powered in the nanosecond that it took to get here.... 🙂
Write-Host ""
Write-Host "Retrieving a list of powered on guests...." -Foregroundcolor Green
Write-Host ""
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Retrieving a list of powered on guests...."
Add-Content $logfile ""
$poweredonguests = Get-VM -Location $cluster | where-object {$_.PowerState -eq "PoweredOn" }

Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Checking to see if vCenter is virtualized"

# Retrieve host info for vCenter
if ($vcenterVMName -ne "NA")
{
    Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) vCenter is indeed virtualized, getting ESXi host hosting vCenter Server"
    $vCenterHost = (Get-VM $vCenterVMName).Host.Name
    Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) $vCenterVMName currently running on $vCenterHost - will process this last"
}
else
{
    Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) vCenter is not virtualized"
}

Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Proceeding with VM PowerOff"
Add-Content $logfile ""

# And now, let's start powering off some guests....
ForEach ( $guest in $poweredonguests )
{
    if ($guest.Name -ne $vCenterVMName)
    {
        Write-Host "Processing $guest.... " -ForegroundColor Green
        Write-Host "Checking for VMware tools install" -Foregroundcolor Green
        $guestinfo = get-view -Id $guest.ID
        if ($guestinfo.config.Tools.ToolsVersion -eq 0)
        {
            Write-Host "No VMware tools detected in $guest , hard power this one" -ForegroundColor Yellow
            Write-Host ""
            Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) $guest - no VMware tools, hard power off"
            Stop-VM $guest -confirm:$false | out-null
        }
        else
        {
           write-host "VMware tools detected.  I will attempt to gracefully shutdown $guest"
           Write-Host ""
           Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) $guest - VMware tools installed, gracefull shutdown"
           $vmshutdown = $guest | shutdown-VMGuest -Confirm:$false | out-null
        }
    }
}

# Let's wait a minute or so for shutdowns to complete
Write-Host ""
Write-Host "Giving VMs 2 minutes before resulting in hard poweroff"
Write-Host ""
Add-Content $logfile ""
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Waiting a couple minutes then hard powering off all remaining VMs"
Sleep 120

# Now, let's go back through again to see if anything is still powered on and shut it down if it is
Write-Host "Beginning Phase 2 - anything left on.... night night...." -ForegroundColor red
Write-Host ""

# Get our list of guests still powered on...

$poweredonguests = Get-VM -Location $cluster | where-object {$_.PowerState -eq "PoweredOn" }

if ($poweredonguests -ne $null)
{
	ForEach ( $guest in $poweredonguests )
	{
		if ($guest.Name -ne $vCenterVMName)
		{
			Write-Host "Processing $guest ...." -ForegroundColor Green
			#no checking for toosl, we just need to blast it down...
			write-host "Shutting down $guest - I don't care, it just needs to be off..." -ForegroundColor Yellow
                        Write-Host ""
			Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) $guest - Hard Power Off"
			Stop-VM $guest -confirm:$false | out-null
		}
	}
}

# Wait 30 seconds
Write-Host "Waiting 30 seconds and then proceding with host power off"
Write-Host ""
Add-Content $logfile ""
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Processing power off of all hosts now"
Sleep 30

# and now its time to slam down the hosts - I've chosen to go by datacenter here but you could put the cluster
# There are some standalone hosts in the datacenter that I would also like to shutdown, those vms are set to
# start and stop with the host, so i can just shut those hosts down and they will take care of the vm shutdown 😉

$esxhosts = Get-VMHost -Location $cluster
foreach ($esxhost in $esxhosts)
{
    if ($esxhost.Name -ne $vCenterHost)
    {
        #Shutem all down
        Write-Host "Shutting down $esxhost" -ForegroundColor Green
        Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Shutting down $esxhost"
        $esxhost | Foreach {Get-View $_.ID} | Foreach {$_.ShutdownHost_Task($TRUE)}
    }
}

# Finally shut down the host which vCenter resides on if using that functionality
if ($vcenterVMName -ne "NA")
{
    Add-Content $logfile ""
    Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) Processing $vcenterHost - shutting down "
    Get-VMhost $vCenterHost | Foreach {Get-View $_.ID} | Foreach {$_.ShutdownHost_Task($TRUE)} | out-null
	Write-Host "Shutdown Complete." -ForegroundColor Green
}

Add-Content $logfile ""
Add-Content $logfile "$(get-date -f dd/MM/yyyy) $(get-date -f HH:mm:ss) All done!"
# That's a wrap

Just to say something obvious, test!

I used vCenter Simulator 2.0 for initial testing and once happy I used a real environment.

Now go and turn off all the vSphere environments you can! (just kidding)

(Seriously, I’m just kidding, don’t do it! Really! Just don’t)

%d bloggers like this: