Major We have multiple slave nodes, on windows, all having the number of executors set to 1.
Our workflow is as follows:
Job 1 requests an executor with a specific label and calls Job 2 (waiting for it's completion)
Job 2 requests another executor with same label and calls Job 3 (waiting for it's completion).
Very seldom, we have runs where the same node (with the number of executors set to 1) gets assigned 2 times (we always assign 2 executors to a run).
[JENKINS-24831] Executor gets assigned multiple times
Octavian Ciuhandu
created issue -
Octavian Ciuhandu
added a comment - I'm attaching 2 screenshots with the definitions of Job 1 and Job 2.
As for plugins, we use: Gearman 0.0.7, Build timeout 1.14, Event Publisher 0.0.3, Node and Label parameter 1.5.1, Parameterized Trigger 2.25, Throttle Concurrent Builds 1.8.3
Octavian Ciuhandu
added a comment - I'm attaching 2 screenshots with the definitions of Job 1 and Job 2.
As for plugins, we use: Gearman 0.0.7, Build timeout 1.14, Event Publisher 0.0.3, Node and Label parameter 1.5.1, Parameterized Trigger 2.25, Throttle Concurrent Builds 1.8.3 Octavian Ciuhandu
added a comment - Job 01 Octavian Ciuhandu
made changes -
| Attachment | New: Screen Shot 2014-09-23 at 22.53.09.png [ 26928 ] |
Octavian Ciuhandu
added a comment - Job 02 Octavian Ciuhandu
made changes -
| Attachment | New: Screen Shot 2014-09-23 at 22.54.26.png [ 26929 ] |
Please explain what you mean here:
gets assigned 2 times (we always assign 2 executors to a run).
The jobs' config.xml would also be helpful.
Why are you passing in a parameter to 02 if it doesn't take parameters?
Octavian Ciuhandu
added a comment - Job 2 calls another job (Job 3) that actually uses both parameters.
Job 1 config.xml: http://paste.openstack.org/show/114710/
Job 2 config.xml: http://paste.openstack.org/show/114711/
By "gets assigned 2 times" I mean that when I check the 2 parameters (that are set in Job 1 and Job 2) they have the same value, e.g.:
hyperv01: node1.domain.tld
hyperv02: node1.domain.tld
Since each job states that the parameter hyperv01/hyperv02 takes value $NODE_NAME and there's only one executor per slave node, this should be impossible.
Octavian Ciuhandu
added a comment - Job 2 calls another job (Job 3) that actually uses both parameters.
Job 1 config.xml: http://paste.openstack.org/show/114710/
Job 2 config.xml: http://paste.openstack.org/show/114711/
By "gets assigned 2 times" I mean that when I check the 2 parameters (that are set in Job 1 and Job 2) they have the same value, e.g.:
hyperv01: node1.domain.tld
hyperv02: node1.domain.tld
Since each job states that the parameter hyperv01/hyperv02 takes value $NODE_NAME and there's only one executor per slave node, this should be impossible. Octavian Ciuhandu
added a comment - As a note, it's something that happens only occasionally, we have seen this happening about around 5 times in the last week or so, and we have not been able to identify any possible reason for this.
Octavian Ciuhandu
added a comment - As a note, it's something that happens only occasionally, we have seen this happening about around 5 times in the last week or so, and we have not been able to identify any possible reason for this.
Could you provide a minimal, complete, verifiable example of jobs that have this problem? Which plugins are used (e.g. to trigger + wait, to require two executors per run, ...), in which versions?