Lab 6.1 - stucked again with node-lab...

laurent.roffe

Hello,

I am still blocked now to reproduce the lab 6.1 of the LFS268 training program

the - very simple - javascript application template will not build in my cluster : https://github.com/LFS268/node-lab

The job is in a pending state forever, and I can see some messages in the pod events, stating that the pod fails to start due to unresolved scheduling in the cluster :

Warning  FailedScheduling   56s (x12 over 9m45s)    default-scheduler   0/3 nodes are available: 3 Insufficient cpu.

At first sight I had tried to upscale my cluster switching node_machine_type from n1-standard-2 to n1-standard-4, but it does not change anything !

This pipeline run from the javascript build pack : https://github.com/jenkins-x/jx3-pipeline-catalog/blob/master/packs/javascript/.lighthouse/jenkins-x/release.yaml translates into a pod to initialize with 10 containers inside. Could it be the problem ? launching 10 containers does not seem to me such a big issue for a GCP cluster of that size (3 nodes, 12 CPUs and 45 Go of RAM)

Thank you for your help and advises

Regards
Laurent

k get pods
NAME                                                              READY   STATUS      RESTARTS   AGE
myrepo-node-lab3-main-release-7cs56-from-build-pack-m6-bn42f   0/10    Pending     0          40m
...
 
---
 
k describe pod myrepo-node-lab3-main-release-7cs56-from-build-pack-m6-bn42f
...
  Warning  FailedScheduling   56s (x12 over 9m45s)    default-scheduler   0/3 nodes are available: 3 Insufficient cpu.
 

You can find the k describe pod result below, but in fact

laurent.roffe

I finally understand what happens : in april Jenkins-x have adapted all the pipelines of the catalog to comply with an upgrade of Tekton release : the PR is here : https://github.com/jenkins-x/jx3-pipeline-catalog/pull/984

It implies that the resources claims for the pod are declared in a step specifically (here jx-variable which is a common step) and not at the stepTemplate level. otherwise each container in the pod will claim the specified amount of resources, which is huge.

But, I don't know why, for javascript the merge is not Ok, so I tested my fix and propsed a pull request to update the catalogue : https://github.com/jenkins-x/jx3-pipeline-catalog/pull/1208

Hope that this PR will be approved and this comment / research can help some of you blocked with this lab

Laurent

laurent.roffe

My pull request above has been approved and merged by the Jenkins X team