-
Improvement
-
Resolution: Duplicate
-
Minor
-
None
-
Jenkins 1.583, Windows OS
I have a multi-configuration Job that has to run on multiple nodes at the same time. I want the job to checkout from CVS, then do build steps on the CVS content.
Current situation
-Multiple slave nodes are assigned to the job.
-Master is NOt assigned to the job.
-CVS is accessed by pserver CVSROOT and defined in the job configuration
Current behaviour
-Master checks out from CVS, then delegates job to the slave nodes, which also do a CVS checkout, then do the job
Expected behaviour
-Master immediately delegates to the assigned nodes without doing a CVS chekout first. Since the master is not an assigned node of the job, it is not supposed to do anything concerning the job.
- duplicates
-
JENKINS-14618 ignore the checkout on the master
-
- Reopened
-
[JENKINS-25040] Unnecessary SCM actions in multi-configuration job on master
In this case JENKINS-14618 is a better match.
Is this about saving time/space, or does master have connectivity issues or something else preventing successful checkout?
Alexander Ziller
added a comment - Yes JENKINS-14618 would fit the description. The nodes of the matrix run on different machines, where the matrix job is to be executed. The master's only job is to distribute the job to the slaves. As Daniel Beck says, the issue JENKINS-4960 describes the exact opposite where the master is supposed to do the checkout.
If I follow the idea of JENKINS-4960, I would indeed have a problem since the master is on a slow network subnet, where-as the slaves have direct access to the SCM. Having the master do the checkout and then somehow transfer the data to the slaves would mean a big increase in performance/time loss.
So yes, as Daniel Beck already stated, JENKINS-14618 describes my point, altho somewhat more compact... JENKINS-4960 is not related to my problem.
I intent to save time by directly passing the job to the slaves configured in the job. The master has no actual work to do except hand the job to the slaves and receive the result of the job.
Alexander Ziller
added a comment - Yes JENKINS-14618 would fit the description. The nodes of the matrix run on different machines, where the matrix job is to be executed. The master's only job is to distribute the job to the slaves. As Daniel Beck says, the issue JENKINS-4960 describes the exact opposite where the master is supposed to do the checkout.
If I follow the idea of JENKINS-4960 , I would indeed have a problem since the master is on a slow network subnet, where-as the slaves have direct access to the SCM. Having the master do the checkout and then somehow transfer the data to the slaves would mean a big increase in performance/time loss.
So yes, as Daniel Beck already stated, JENKINS-14618 describes my point, altho somewhat more compact... JENKINS-4960 is not related to my problem.
I intent to save time by directly passing the job to the slaves configured in the job. The master has no actual work to do except hand the job to the slaves and receive the result of the job. Duplicates JENKINS-14618. Resolving this one as the other is older and has more watchers. Feel free to copy the description why you need this over into a 'Me, too' comment.
Alexander Ziller
added a comment - Issue is duplicate. Necessary information has been copied to comment at primary issue.
Alexander Ziller
added a comment - Issue is duplicate. Necessary information has been copied to comment at primary issue. 
Your request seems to be similar to JENKINS-4960.
Would it resolve your case?