I'm dealing with a very weird anomaly in Jenkins that makes absolutely no sense to me. Essentially Jenkins is behaving differently for a powershell command than for a batch command.
My goal is to pass an environment variable (or parameter) from one Jenkins Job to another. However this variable to be passed is generated during the runtime of the first job.
I made a fake project to test passing variables and I was able to do so by adding a build step to echo out the variable into an env.props file on the node and then used the parameterized trigger plugin to call the next job. I was able to get this to work great in this test scenario but when I tried to implement the same steps in the actual build job (which relies on powershell scripts) it did not work.
After, a lot of trial and error I have found that when I use a windows batch command to echo the variable into a props file and then inject the variable into the job - it works perfect. But when I do the exact same thing with a powershell command it does not inject the variable back in to the job even though I use the exact same line of code. It still writes the variable to the file but Jenkins will not "reinject" this variable back into the job's env variable even though I am using the exact same step to do so.
The command is essentially this:
echo Testvar=Somevalue > C:\Jenkins\env.props
Both sucesffuly write the string to the props file, but when done with a powershell command, Jenkins will not absorb the txt from the run. Almost as if powershell is encoding it in a way that Jenkins cannot read but looks the exact same to me.
Any ideas?
Turns out, it was the encoding!
echo "string" > file.txt
does not produce the same result in batch as powershell.
Switching to
echo "string" | out-file -encoding ASCII file.txt
did the trick.
Related
I am trying to modify several values of a PowerShell script template prior to execution. The approach I took was to use the Get-Content command to read the template file, then to use the replace operand to replace the content with a content of my choosing, and then the Set-Content command to update the file, then execute the script. However, according to the error messages that I encounter, it seems that the modified file is not running, but the template one.
The thing I find hard to understand is why, as I use the Get-Content once again and print the result prior to execution, where I witness that the file did change.
I use simple PowerShell scripting with no threads or so ever, the script is being executed on an Azure VM via the Run-Command feature. I wonder why it happens.
Can anyone please explain?
I created a freestyle job in Jenkins that I just set up (latest version).
I added parameters to it. One of those is a Options selection for ReleaseType with the options of Staging and Release.
One of the build steps is executing a remote command on the server when the site is uploaded to. It uses the Execute Windows Batch Command build step.
Here is the command line (with things made generic):
sexec myuser#mysite.com -pw=mypassword -cmd="PowerShell -Command ""C:\batch\bvCopyFast.ps1 C:\inetpub\mysite${ReleaseType}\siteLoad C:\inetpub\mysite${ReleaseType}\site""
Basically I am executing a powershell command that uses Robocopy to copy the files from the upload folder to the actual release folder for the site.
As you can see I need to have the ${ReleaseType} replaced with the actual value. The problem is that when this gets executed it isn't doing the substitution. I just uses that literal value in the command and that doesn't work.
If you use the -Command parameter it implies you are going to write raw PowerShell code in between the quotation marks that follow (allow you can call a script as you have).
PowerShell -Command "Get-Date; pause;"
To call a PowerShell script file you should use:
PowerShell -File "Your-Script.ps1 -Parameter1 Argument1 -Parameter2 Argument2"
https://learn.microsoft.com/en-us/powershell/scripting/components/console/powershell.exe-command-line-help?view=powershell-6
I would write a PowerShell script that accepted your root path and the releaseType as arguments and execute that.
Param($rootPath,$releaseType)
{
robocopy "$($rootPath)\$($releaseType)\siteLoad" "$($rootPath)\$($releaseType)\site"
}
I have never used Jenkins so I hope this works as I expect it to!
sexec myuser#mysite.com -pw=mypassword -cmd=""PowerShell -File 'C:\batch\newScript.ps1' -RootPath 'c:\inetpub\mysite' -ReleaseType {ReleaseType}""
Is it possible to call a PowerShell command directly in the pipelines groovy script? While using custom jobs in Jenkins I am able to call the command with the PowerShell Plugin. But there is no snippet to use this in the groovy script.
I also tried sh() but it seems that this command does not allow multiple lines and comments inside the command.
To call a PowerShell script from the Groovy-Script:
you have to use the bat command.
After that, you have to be sure that the Error Code (errorlevel) variable will be correctly returned (EXIT 1 should resulting in a FAILED job).
Last, to be compatible with the PowerShell-Plugin, you have to be sure that $LastExitCode will be considered.
I have notice that the 'powershell' is now available in pipeline, but since it have several issues I prefer this variant. Still waiting it works stabil. I actually have an issue with the 'dontKillMe' behavior.
Since Jenkins 2.207 with Powershell plugin 1.4, I have replace all my calls with the official powershell pipeline command. I do now recommend to use it.
Note that you must predent \$ErrorActionPreference='Stop'; to your Script if you want it to abort on Write-Error because of an Issue with the powershell plugin.
For that porpuse I have written a little groovy method which could be integrate in any pipeline-script:
def PowerShell(psCmd) {
psCmd=psCmd.replaceAll("%", "%%")
bat "powershell.exe -NonInteractive -ExecutionPolicy Bypass -Command \"\$ErrorActionPreference='Stop';[Console]::OutputEncoding=[System.Text.Encoding]::UTF8;$psCmd;EXIT \$global:LastExitCode\""
}
[EDIT] I have added the UTF8 OutputEncoding: works great with Server 2016 and Win10.[/EDIT]
[EDIT] I have added the '%' mask[/EDIT]
In your Pipeline-Script you could then call your Script like this:
stage ('Call Powershell Script')
{
node ('MyWindowsSlave') {
PowerShell(". '.\\disk-usage.ps1'")
}
}
The best thing with that method, is that you may call CmdLet without having to do this in the Script, which is best-praxis.
Call ps1 to define CmdLet, an then call the CmdLet
PowerShell(". '.\\disk-usage.ps1'; du -Verbose")
Do not forget to use withEnv() an then you are better than fully compatible with the Powershell plugin.
postpone your Script with . to be sure your step failed when the script return an error code (should be preferred), use & if you don't care about it.
Calling PowerShell scripts is now supported with powershell step as announced on Jenkins blog.
The documentation mentions it supports multiple lines scripts.
From version 2.28 of Pipeline Nodes and Processes Plugin, we can directly use 'powershell'.
Eg: powershell(". '.Test.ps1'")
You can use the sh command like this:
sh """
echo 'foo'
# bar
echo 'hello'
"""
Comments are supported in here.
I have a Jenkins (1.493) project that uses the Jenkins PowerShell build step to execute a PowerShell script. Inside that script I want to invoke another script that is stored inside a file. I have now reduced it to the following:
Script inside Jenkins PowerShell build step:
& "\\stemmer.local\sidevelopment\cvdev\devbase\jenkins\PowerShell\Test.ps1"
Content of Test.ps1:
write-host 'Hello world!'
Whenever this Jenkins project executes, the PowerShell build step hangs indefinitely.
Things I have tried/verified so far:
Adding some output before the invocation of Test.ps1 shows me that the Jenkins PowerShell script is being execute normally up to the point where Test.ps1 is called.
The file Test.ps1 exists and is reachable from the build slave that executes the script. If I alter the file's name, I get the expected error message from PowerShell...
Exchanging the " for ' in the 1st script does not change anything. Also, using dot-sourcing rather than & does not help.
The file Test.ps1 can be executed properly from the powershell itself using the same command line that is being used in the Jenkins PowerShell script.
The execution policy for PowerShell has been set to unrestricted on my development host as well as on the Jenkins build slave.
I've tried replacing the PowerShell build step with a Windows batch command build step that looks like this:powershell.exe -InputFormat None -File "\\stemmer.local\sidevelopment\cvdev\devbase\jenkins\PowerShell\Test.ps1"and played around a little with the parameters of powershell.exe, but the results were - in those cases that were syntactically and otherwise correct as far as I can tell - always the same.
I only found few references to problems that sounded similar, but none of the approaches mentioned elsewhere did help me fix this. I am absolutely puzzled, and wondering whether someone encountered this issue before (and maybe even got a scenario like the one I have in mind running).
Thanks a lot for any input!
Volker
have you tried to set execution policy to bypass ?
Copy the script file locally, then invoke it from within the Jenkins PowerShell plugin - that way it works as expected.
I'm trying to set up a new build configuration in TeamCity using the Powershell runner. However, I can't seem to find a way to access the TeamCity System Properties in the build script. I've seen hints that it is possible, but cannot find documentation on how to do it.
I have tried accessing the system properties using Powershell variable syntax, $variable. I have also printed out all variables in memory and see no teamcity variables to use.
Is this possible with the Powershell runner, and if so what is the syntax necessary to get it working?
TeamCity will set up environment variables, such as build.number (you can see a list of these within TeamCity).
In Powershell you can access environment variables using the env "provider", e.g.
$env:PATH
TeamCity variables are accessible by replacing the . with a _, so the build.number variable can be accessed as
$env:build_number
As it says in the TeamCity documentation, the system parameters are passed to the build script runner, but not all build script runners know what to do with them. In the case of the Powershell script runner, when using a script file, they don't propagate down to your scripts.
It's occurred to me to write a psake-optimized build runner that does, but in the meantime you can do one of the following:
explicitly map any of the TeamCity build properties to script parameters using the parameter expansion that's available within the Script Source box. eg .\build.ps1 -someParam:%build.name%
use environment parameters, which can be accessed explicitly within PowerShell using $env:NAME_IN_TEAMCITY syntax, eg $env:TEAMCITY_VERSION, or looped over and pushed into variable scope
access the build properties file that TeamCity makes available during the build. The file is available at $env:TEAMCITY_BUILD_PROPERTIES_FILE, and if you load the XML version it's fairly easy to loop through and push them all into scope (though you do get everything as a string of course). I posted a gist on how to do this (https://gist.github.com/piers7/6432985). Or, if using Psake, modify the script above to return you a hashtable which you can pass directly to Psake's -properties argument.
It is posible. Here is example how to pass system properties into PSake script:
& .\psake.ps1 -parameters #{build_number=%build.number%; personal_build=%build.is.personal%}
If you don't use Psake, you can define your variables like this:
$build_number = %build.number%
The %build.number% part will be replaced with TeamCity provided data. I think, it works only in Source code script input mode.
I created a meta-runner that will pass through System parameters to parameters declared in the Powershell script. It's not perfect (if you put '# in your source it will break) but it works for what I needed, you can find it here: https://gist.github.com/anonymous/ef60ada3f48f0fb25093