Now that I know both Jenkins/jjb and OpenShift, I find myself often thinking in terms of *services* as opposed to the default "job" model of Jenkins.
Basically, a number of the things I'd like to do in the CI env would be better expressed as services than as jobs. (Not all)
I heard there was an effort to create an internal OpenShift environment - are there any examples of consumers of that or documentation on it?
Are people using that for *persistent* services or just testing container images?
But even then I'd really like to expose public services in some cases - think informational display that gets awkward to render into jobs.
I have access to other OpenShift clusters where I could run these services, but then I get into a situation of copying data around, as well as dealing with authentication.
(IMO really CentOS CI should have *multiple* Jenkins, each group gets its own master deployed *in* OpenShift...a while ago I accidentally forgot a `node:` spec on one of my jobs which ended up running on the libvirt FreeBSD slave. This is something that should really never happen)
On Mar 03 14:39, Colin Walters wrote:
Now that I know both Jenkins/jjb and OpenShift, I find myself often thinking in terms of *services* as opposed to the default "job" model of Jenkins.
Basically, a number of the things I'd like to do in the CI env would be better expressed as services than as jobs. (Not all)
I heard there was an effort to create an internal OpenShift environment - are there any examples of consumers of that or documentation on it?
This is still in the bringup stage.
Are people using that for *persistent* services or just testing container images?
"Yes" :) There are a few persistent services that are likely to end up in CI Openshift, but testing container images is a goal as well.
But even then I'd really like to expose public services in some cases - think informational display that gets awkward to render into jobs.
I have access to other OpenShift clusters where I could run these services, but then I get into a situation of copying data around, as well as dealing with authentication.
(IMO really CentOS CI should have *multiple* Jenkins, each group gets its own master deployed *in* OpenShift...a while ago I accidentally forgot a `node:` spec on one of my jobs which ended up running on the libvirt FreeBSD slave. This is something that should really never happen)
I think this scenario is about to become "painful" (we knew/know this is coming). The reason for a single master was that the Jenkins frontend was meant to be the point of collaboration to visualize the 'matrix' of tests between related projects but lately it's more noisy than anything.
Cheers! --Brian
On Fri, Mar 3, 2017, at 02:49 PM, Brian Stinson wrote:
This is still in the bringup stage.
Is it on baremetal? This would make it easier to do virt-in-containers.
If you recall I use Duffy to allocate worker machines which may do builds/tests, but not necessarily *on* or for the base CentOS 7 host.
(BTW, I'd still like Atomic Host as a provisioning option)
I've been thinking about changing this to set up an OpenShift cluster and have the jobs schedule containers into it...but the awkward thing about this (in thought experiment stage, not getting into things like networking across nodes) would be the duffy timeout killing the masters.
I think this scenario is about to become "painful" (we knew/know this is coming). The reason for a single master was that the Jenkins frontend was meant to be the point of collaboration to visualize the 'matrix' of tests between related projects but lately it's more noisy than anything.
Yeah. I really don't like any of the Jenkins web UI to be honest, though the "blue ocean" stuff for Pipeline looks decent. But that's more of a targeted specific case.
On Fri, Mar 3, 2017 at 3:11 PM, Colin Walters walters@verbum.org wrote:
On Fri, Mar 3, 2017, at 02:49 PM, Brian Stinson wrote:
This is still in the bringup stage.
Is it on baremetal? This would make it easier to do virt-in-containers.
If you recall I use Duffy to allocate worker machines which may do builds/tests, but not necessarily *on* or for the base CentOS 7 host.
(BTW, I'd still like Atomic Host as a provisioning option)
I've been thinking about changing this to set up an OpenShift cluster and have the jobs schedule containers into it...but the awkward thing about this (in thought experiment stage, not getting into things like networking across nodes) would be the duffy timeout killing the masters.
I think this scenario is about to become "painful" (we knew/know this is coming). The reason for a single master was that the Jenkins frontend was meant to be the point of collaboration to visualize the 'matrix' of tests between related projects but lately it's more noisy than anything.
Yeah. I really don't like any of the Jenkins web UI to be honest, though the "blue ocean" stuff for Pipeline looks decent. But that's more of a targeted specific case.
I think the Openshift/Jenkins/Jenkins Pipeline is the long term goal and whether you use the Openshift or Jenkins Blue Ocean UI doesn't matter since the information is presented in both.
Ci-users mailing list Ci-users@centos.org https://lists.centos.org/mailman/listinfo/ci-users
-
Ari LiVigni -
Brian Stinson -
Colin Walters