Jx: Pipeline hung waiting because maven is offline

[harpreetsingh]

Steps to reproduce:

• Install jenkins-x on a new gke cluster through cluster create;
• jx create spring -d web -d actuator; command returned
  Opened jenkins job and found that it had hung.

[rawlingsj]

I wonder if it did eventually come up? Starting a maven build pod means the node kubernetes schedules it to run on needs to pull the image which can take some time but on GKE only a couple of mins max.

So perhaps it's something else, maybe there's not enough resources in the cluster to schedule the maven build pod. Right now there's no easy way to find out other than using the maven pod name at the bottom of the log in the screenshot or running:

kubectl get pods

find the maven build pod

kubectl describe pod [maven build pod id]

This should show some 'events' that should explain why the pod isn't starting.

__notes__ we need to provide an easy way to know why a build pod cant start. Even if to start with it's a link in the jenkins console log to an FaQ.

[rawlingsj]

as a side note a big +1 for the issue - thanks @harpreetsingh

[jstrachan]

agreed - we need some way to warn the user if a pipeline can't start.

Incidentally as part of #412 the jx create spring, jx create quickstart and jx import commands now output the info text:

Note that your first pipeline may take a few minutes to start while the necessary docker images get downloaded!

as even when everything is working perfectly, it can take a few minutes for the first pipeline to start.

[jstrachan]

jx logs jenkins should show the jenkins logs which might include any errors if build pods cannot be created - e.g. due to lack of resources or underlying kubernetes issues

[pax95]

Have the same hanging maven build pod when running jx on minikube.
kubectl describe pod maven-qx7hv reports ->

Events:
  Type     Reason            Age               From               Message
  ----     ------            ----              ----               -------
  Warning  FailedScheduling  3s (x7 over 34s)  default-scheduler  0/1 nodes are available: 1 Insufficient cpu.

If thats the root cause it would be nice to surface the problem using jx

[rawlingsj]

We have a jx status command that helps identify resource issues too BTW. Ideally we'd tie in prometheus alerts and get notifications for low resources too.

[rawlingsj]

I think this can be closed now. The recommended initial cluster size is better now, we also enable cluster autoscaling where possible and provide a jx status command to help identify this situation. Please reopen if there's more we can do here.