A while ago I posted two entries
- Invoke script blocks locally or remotely where I proposed an enhanced version of
Invoke-Command
. - Code as configuration where I explain the basics of the concept.
With [SDL] I try to establish a no login culture. That is we should not have to Remote Desktop into a target computer but we should do everything from our client machine.
With PowerShell that is easy and you can execute script blocks on a remote target with Invoke-Command
or you establish a remote session to the target with Enter-PSSession
.
When automating, Invoke-Command
becomes the focus point because a remote session cannot execute scripts.
Let’s build up the case. Let’s hypothesize that I want to execute this script block locally or on a remote target $ComputerName
.
$scriptBlock1= {
Write-Host "$($env:COMPUTERNAME) says hello"
}
if ($ComputerName)
{
Invoke-Command -ComputerName $ComputerName -ScriptBlock $scriptBlock1
}
else
{
Invoke-Command -ScriptBlock $scriptBlock1
}
A while ago, I’ve also posted the Code as configuration where I explain the basics of the concept.
A key requirement is that the scripts need to be as clean as possible. The conditional invocation of the $scriptBlock1
is not the best example of clean.
This becomes more complicated as the number of script blocks and script files increase.
For this reason a while ago I posted the Invoke script blocks locally or remotely where an enhanced version of Invoke-Command
can do this
$scriptBlock1= {
Write-Host "$($env:COMPUTERNAME) says hello"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlock1
This is much better. One line invokes a script block regardless of the value of $ComputerName
. This is simple and clean and complies with the basics of code as configuration.
As I started building scripts another problem started to become apparent. Let’s assume I want to improve the script and I want to control what the script block says in Write-Host "$($env:COMPUTERNAME) says hello"
There are two options here:
- Add script block parameters
- Use script scope parameters directly.
For example
#region example with script block parameter
$scriptBlockWithParam= {
param($Message)
Write-Host "$($env:COMPUTERNAME) says hello"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlockWithParam -ArgumentList "Hello"
#endregion
#region example with direct reference to script variables
$message="Hello"
$scriptBlockWithVariable= {
Write-Host "$($env:COMPUTERNAME) says $message"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlockWithVariable
#endregion
The second example works only when $ComputerName
is $null
. When a script block executes remotely then the variables in the script do not flow into the remote execution context.
To make it work we need to use the $Using:variableName
assignment like this
#region example with direct reference to script variables
$message="Hello"
$scriptBlockWithVariable= {
Write-Host "$($env:COMPUTERNAME) says $($Using:message)"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlockWithVariable
#endregion
This will work remotely but not locally because $Using
is not allowed in non-remote invocation. It turns out that the same principal applies when feeding the script blocks to PowerShell jobs.
To fix the script block we need a conditional variable assignment. PowerShell offers the $PSSenderInfo
variable when a block executes remotely. When not then it is $null
.
The script becomes
$message="Hello"
$scriptBlockWithVariable= {
if($PSSenderInfo) {
$message=$Using:message
}
Write-Host "$($env:COMPUTERNAME) says $($Using:message)"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlockWithVariable
The above works with all cases. You might be wondering why I focus on this paradigm and not the one with the parameterized script block. I have two reasons:
- When trying to create scripts for code as configuration you need to keep the script as clean and easy to understand as possible. The definition of parameters in the script block and the additional reference in the invocation line do not comply. A script block is like a function. Very useful for executing multiple times with some variations but in code as configuration that is not the goal. The only reason we need a script block is to make a remote call. The variations will be expressed in different cmdlets within the script block body.
- I like promoting my session’s
$DebugPreference
and$VerbosePreference
values into the remote execution context. As with any other parameters, the local values do not automatically flow into the remote executions so I need to resort to conditional$PSSenderInfo
assignment as mentioned above. I need to do this regardless of my implementation flavour, that is using script block parameters or not.
Let’s see how it looks
$message="Hello"
$scriptBlockWithVariable= {
if($PSSenderInfo) {
$DebugPreference=$Using:DebugPreference
$VerbosePreference=$Using:VerbosePreference
}
if($PSSenderInfo) {
$message=$Using:message
}
Write-Debug "Going to say something"
Write-Host "$($env:COMPUTERNAME) says $($Using:message)"
Write-Verbose "Said something"
}
Invoke-CommandWrap -ComputerName $ComputerName -ScriptBlock $scriptBlockWithVariable
The last three lines represent my code as configuration blocks. This is where I get to execute multiple cmdlets which internally output Debug and Verbose messages. It actually looks more clean than this but at this moment I can’t show an example yet. The rest of the block is just noise. Imagine you have multiple script blocks in one script file. Then most of my script is focused in facilitating the remote call and not the actual code as configuration parts. I don’t need this and it’s too noisy and complicated. Also, it doesn’t help people to copy paste examples into the file, without having deep knowledge of what makes remoting tick.
So I enhanced even more the Invoke-CommandWrap
and I updated the online gist.
- When the script will execute remotely then the cmdlet will
- Create a new array of script block
- Add the section for the
$DebugPreference
and$VerbosePreference
in the array. - When optional parameter
-UseParameters
is defined then add a section with script variables that are referenced in the array. - Add the originally script block in the array.
- Merge the array items into a new script block and replace the original value.
- Brand the execution of the script block. Added optional parameter
-BlockName
. This is a personal flavor for logging purposes. It is not actually required.
The above script is simplified into this:
$message="Hello"
$scriptBlock= {
Write-Debug "Going to say something"
Write-Host "$($env:COMPUTERNAME) says $($Using:message)"
Write-Verbose "Said something"
}
Invoke-CommandWrap -BlockName 'Talk' -ComputerName $ComputerName -ScriptBlock $scriptBlock -UseParameters @("message")
Now my script block is pure and clean. It contains only the lines that are relevant to my intention.
The only ugly thing is that I need to remember to reference the dependent parameters in the invocation. Please notice the extra -UseParameters @("message")
.
What happened internally?
- If the
$ComputerName
is$null
then nothing happened. This is really nice because I still get to debug my script block. - If the
$ComputerName
has a value then internally the value fed to the-ScriptBlock
parameter was modified. Because the original block was replaced, debug capability is lost but to be honest, since I could never debug remote script blocks through PowerShell ISE, I don’t see the problem. You can actually see the new generated body for the script block that is going to execute when the$DebugPreference
isContinue
.
It is possible to skip the necessity for -UseParameters @("message")
by promoting all script variables in the generated script block.
But at this point I don’t feel comfortable with the idea. For this reason I would appreciate any feedback you can provide.
Leave a Comment