-
Bug
-
Resolution: Fixed
-
Major
-
-
1.27.1.1
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing
Reproduction:
add at least 2 pod templates
set a concurrency limit of 1 on the first and a higher one on the other
concurrency for the whole cluster above the two limits
spawn some pods of the second template
-> a job requiring a pod of template one does not spawn until no other pods are live
- is caused by
-
-
- Reopened
-
[JENKINS-63705] Concurrency limit not calculated over template anymore in 1.27.1
Sledge Sulaweyo
created issue -
Sledge Sulaweyo
made changes -
| Description | Original: With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing |
New:
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing Reproduction: ad at least 2 pod templates set a concurrency limit of 1 on the first and a higher one on the other concurrency for cluster above the two limits spawn pods of the second template -> a job requiring a pod of template one does not spawn |
Sledge Sulaweyo
made changes -
| Description |
Original:
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing Reproduction: ad at least 2 pod templates set a concurrency limit of 1 on the first and a higher one on the other concurrency for cluster above the two limits spawn pods of the second template -> a job requiring a pod of template one does not spawn |
New:
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing Reproduction: ad at least 2 pod templates set a concurrency limit of 1 on the first and a higher one on the other concurrency for cluster above the two limits spawn pods of the second template -> a job requiring a pod of template one does not spawn until no other pods are live |
Sledge Sulaweyo
made changes -
| Description |
Original:
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing Reproduction: ad at least 2 pod templates set a concurrency limit of 1 on the first and a higher one on the other concurrency for cluster above the two limits spawn pods of the second template -> a job requiring a pod of template one does not spawn until no other pods are live |
New:
With the latest release, most likely since JENKINS-59959, the concurrency limit is no longer checked against the pod templates count. I have multiple pod templates in my environment and in 1.26.4 the concurrency limit is correctly calculated per template - in 1.27.1 i don't get my pods spawned anymore as the limit of the template is checked against the total number of pods. At least that's how it looks to me after a lot of testing Reproduction: add at least 2 pod templates set a concurrency limit of 1 on the first and a higher one on the other concurrency for the whole cluster above the two limits spawn some pods of the second template -> a job requiring a pod of template one does not spawn until no other pods are live |
| Assignee | New: David Schott [ shott85 ] |
https://github.com/jenkinsci/kubernetes-plugin/pull/844 also filed which reverts the PR that introduced this new bug.
https://ci.jenkins.io/job/Plugins/job/kubernetes-plugin/job/1.27.1.x/ branch has been cut to release 1.27.1.1 which reverts https://github.com/jenkinsci/kubernetes-plugin/pull/824 cc jglick_bot
| Link | New: This issue is caused by JENKINS-59959 [ JENKINS-59959 ] |
https://github.com/jenkinsci/kubernetes-plugin/pull/843
is an attempt to craft an emergency fix.