Jenkins pipeline sh does not seem to respect pipe

问题:

I am using a Jenkinsfile in a pipeline on version 2.32.2.

For various reasons I want to extract the version string from the pom. I was hoping I wouldn't have to add the maven help plugin and use evaluate.

I quickly came up with a little sed expression to get it out of the pom which uses pipes and works on the commandline in the jenkins workspace on the executor.

$ sed -n '/<version>/,/<version/p' pom.xml | head -1 | sed 's/[[:blank:]]*<\/*version>//g' 1.0.0-SNAPSHOT

It could probably be optimized, but I want to understand why the pipeline seems to be failing on piped sh commands. I've played with various string formats and am currently using a dollar slashy string.

The pipeline step looks like the following to allow for easy output of the command string:

script {
    def ver_script = $/sed -n '/<version>/,/<version/p' pom.xml | head -1 | sed 's/[[:blank:]]*<\/*version>//g'/$
    echo "${ver_script}"
    POM_VERSION = sh(script: "${ver_script}", returnStdout: true)
    echo "${POM_VERSION}"
}

When run in the jenkins pipeline I get the following console output where it seems to be separating the piped commands into separate commands:

[Pipeline] script
[Pipeline] {
[Pipeline] echo
sed -n '/<version>/,/<version/p' pom.xml | head -1 | sed 's/[[:blank:]]*<\/*version>//g'
[Pipeline] sh
[FRA-198-versioned-artifacts-44SD6DBQOGOI54UEF7NYE4ECARE7RMF7VQYXDPBVFOHS5CMSTFLA] Running shell script
+ sed -n /<version>/,/<version/p pom.xml
+ head -1
+ sed s/[[:blank:]]*<\/*version>//g
sed: couldn't write 89 items to stdout: Broken pipe
[Pipeline] }
[Pipeline] // script

Any guidance out there on how to properly use piped commands in a jenkinsfile ?

回答1:

I finally put some thought into it and realized that pipe subshells are probably causing the issue. I know some of the evils of eval but I ended up wrappping this in an eval:

script {
    def ver_script = $/eval "sed -n '/<version>/,/<version/p' pom.xml | head -1 | sed 's/[[:blank:]]*<\/*version>//g'"/$
    echo "${ver_script}"
    POM_VERSION = sh(script: "${ver_script}", returnStdout: true)
    echo "${POM_VERSION}"
}   

回答2:

If your environment allows it, I've found a simple solution to this problem to be to place your script containing pipes into a file, and then run that with sh, like so:

script.sh:

#!/bin/sh
kubectl exec --container bla -i $(kubectl get pods | awk '/foo-/{ print $1 }') -- php /code/dostuff

Jenkinsfile:

stage('Run script with pipes') {
  steps {
    sh "./script.sh"
  }
}

回答3:

I am also struggling with the usage of pipe inside my jenkins pipeline but as a side note, if you want a simple way to extract the version of a maven pom, here's a very clean one I found in another post and that I'm using :

stage('Preparation') {
 version = getVersion()
 print "version : " + version
}
def getVersion() {
  def matcher = readFile('pom.xml') =~ '<version>(.+)</version>'
  matcher ? matcher[0][1] : null
}

gives you :

[Pipeline] echo
releaseVersion : 0.1.24
[Pipeline] sh

回答4:

The pipeline-utility-steps plugin nowadays includes a readMavenPom step, which allows to access the version as follows:

version = readMavenPom.getVersion()