-
Bug
-
Resolution: Fixed
-
Minor
-
Jenkins: 2.346.3
OS: Linux - 4.14.203-156.332.amzn2.x86_64
workflow-durable-task-step:1217.v38306d8fa_b_5c
I thought I saw this fixed in 1210.va_1e5d77e122b
but we are hitting it in 1217.v38306d8fa_b_5c
We hit this in one of our libraries, and were able to refactor around it. I've attached our Jenkins info, and our minimal Jenkinsfile reproduction.
From our console output, I distilled what we were seeing on our AWS node agent.
"start workspace : workspace is /home/ec2-user/workspace/enkins-reproworkspacebug2_master to dir() : /home/ec2-user/workspace/enkins-reproworkspacebug2_master/vars to new node() and workspace : workspace is /home/ec2-user/workspace/enkins-reproworkspacebug2_master__2 BUT cwd() is old workspace/dir : /home/ec2-user/workspace/enkins-reproworkspacebug2_master/vars"
This looks related to https://issues.jenkins.io/browse/JENKINS-69936 but it does not seem like it was fixed for us? Or perhaps it is subtly different.
- is caused by
-
JENKINS-49707 Auto retry for elastic agents after channel closure
-
- Resolved
-
- is related to
-
JENKINS-70693 local workspace is used instead of external workspace
-
- Resolved
-
- relates to
-
JENKINS-69936 PWD returning wrong path
-
- Resolved
-
-
JENKINS-49651 Extend plugin/update center metadata with known incompatibilities
-
- Open
-
- links to
It should be noted that if the node() call was to a different host, we would not see this issue. The issue was that if we node() to the same host (either by accident, or label maps to the same), we DO get a new WORKSPACE, but we DON'T get a correct cwd()