| The lock step can obtain a null resource after saving a global system configuration by clicking on "Manage Jenkins" -> "Configure System" -> "Save". The bug occurs irregardless of any actual changes has been made. steps to reproduce setup
- 4 jobs with the following configuration:
node {
boolean retry = true;
while (retry) {
lock(label: 'res1', quantity: 1, variable: 'RES_NAME') {
println env.RES_NAME
if ("${env.RES_NAME}" != "null") {
retry = false;
sleep(100)
}
}
}
}
- 2 resources configured "r1" and "r2", both with label "res1"
Steps
- Activate a build for each of the jobs simultaneously. 2 of the builds will obtain a lock and then proceed to sleep. 2 of the builds will be blocked waiting for a resource.
- Update the system configuration by clicking on "Manage Jenkins" -> "Configure System" -> "Save"
- One of the blocked build will obtain a "null" resource and re-trying a lock will obtain a resource that is already being used by one of the sleeping build. Demonstrated by the following snippet of the console output:
Running on Jenkins in /var/jenkins_home/workspace/test4@2
[Pipeline] {
[Pipeline] lock
Trying to acquire lock on [Label: res1, Quantity: 1]
Found 0 available resource(s). Waiting for correct amount: 1.
[Label: res1, Quantity: 1] is locked, waiting...
Lock acquired on [Label: res1, Quantity: 1]
[Pipeline] {
[Pipeline] echo
null
[Pipeline] }
Lock released on resource [Label: res1, Quantity: 1]
[Pipeline] // lock
[Pipeline] lock
Trying to acquire lock on [Label: res1, Quantity: 1]
Lock acquired on [Label: res1, Quantity: 1]
[Pipeline] {
[Pipeline] echo
r1
[Pipeline] sleep
Sleeping for 1 min 40 sec[Pipeline] }
Lock released on resource [Label: res1, Quantity: 1]
[Pipeline] // lock
[Pipeline] }
[Pipeline] // node
[Pipeline] End of Pipeline
|
