Yes, the s3Upload portion runs on build slaves.

When used with file parameter, s3Upload on slave works as expected, however.

Thanks for the info. This makes debugging a lot easier. So I can narrow it down to master/slave code.

Yes, I can confirm that it does work on the master for me.

This seems to be related to https://issues.jenkins-ci.org/browse/JENKINS-44000 where aws identity is inherited from master and not from running instance's.

Looking at S3 plugin https://wiki.jenkins.io/display/JENKINS/S3+Plugin, callable is used to stream from slave to master is used as a workaround to upload to s3 since plugin is executed only in master.

I've also tried wrapping step with withAws using string values produced by slave nodes as credential parameters, however this doesn't seem to work as it has to apply to entire scope of pipeline.

Any workaround or solution to this would be wonderful as only viable solution to this is to use shell block with awscli on slave nodes (making this or any aws plugin mute) or run everything on master with every security permission currently only assigned to slave nodes.

Sorry, still no progress here. I am setting up a Slave setup in the next days to replicate this issue.

Hi, I'm facing the same issue in a master/slave setup.

Something peculiar I noticed while goofing around with the settings: if I set the path parameter to be an actual file name (instead of a directory), then the matched files do seem to be uploaded, but obviously they all have the provided file name.

I can confirm this bug, mainly the `path` is fully ignored when include pathPattern is a pattern, making impossible to upload packages to specify target locations inside the buckets. 

I am still trying to find a way to workaround this bug but so solution yet. Anyone?

The only thing that works for me is
 

s3Upload( 
  bucket: 'BUCKET', 
  path: "PATH_TO_FOLDER", // no trailing slash 
  file: "FOLDER", 
  workingDir: "PARENT_OF_FOLDER" 
)

 
With includePath, the only one that worked partially was using "*/.yaml" as path; other patterns like ".yaml", "/" did not work. I say partially because it uploaded only one file eventhough there were several (there is a bug related to this).

Also findFiles is an option, as documented https://github.com/jenkinsci/pipeline-aws-plugin/issues/83.

Sad to see no update on this.

I can confirm it's an issue for me and yes, only on my slaves.

my workaround is uploading file by file

pipeline {
  agent any
    stages {
      stage('Setting up environment variables'){
        def AWS_ACCOUNT_ID = '01010101010101010'
        def REGION = 'eu-north-1'
        def ROLE = 'MyIamRole'
        def EXTERNAL_ID = 'MyExternalId'
        def BUCKET = 'my-artifacts'
        def PROJECT = 'my-project'
      }
      stage ('Build app and upload artifacts to S3'){
        agent {
          label 'my-slave-with-maven'
        }
        steps {
          // build source code
          dir('./SourceCode') {
            sh 'mvn -B clean package'            
          }
        }
        script {
          // upload files to S3
          def jar_files = findFiles(glob: "**/SourceCode/${PROJECT}/target/*.jar")
          jar_files.each {
            echo "JAR found: ${it}"
            withAWS(externalId: "${EXTERNAL_ID}", region: "${REGION}", role: "${ROLE}", roleAccount: "${AWS_ACCOUNT_ID}") {
              s3Upload(file: "${it}", bucket: "${BUCKET}", path: "${PROJECT}/", acl: 'BucketOwnerFullControl')
            }
          }
        }
      }
    }
  }
}