Description of problem:
Only 1 node was close to being out of disk space(had ~200megs free) all the others had ~7GB free. So if it really was the issue the bug should be open against the scheduler to account for disk space when selecting the node.
(from Wesley Hearn)
> > Here is the end of the output about the builder pod:
> >
> > $ oc get pods nodejs-example-3-build -o json
> >
> > "status": {
> > "phase": "Failed",
> > "message": "Pod cannot be started due to lack of disk space.",
> > "startTime": "2015-07-28T20:49:57Z"
> > }
> >
> > Good ole out of disk space.
> >
> > To be clear, build 1 failed clearly. Builds 2 & 3 failed but are reported
> > as succeeded.
> >
> > $ oc get builds
> > NAME TYPE STATUS POD
> > nodejs-example-1 Source Failed nodejs-example-1-build
> > nodejs-example-2 Source Complete nodejs-example-2-build
> > nodejs-example-3 Source Complete nodejs-example-3-build
> >
This is running on https://console.stg.openshift.com/console/
Version-Release number of selected component (if applicable): 3.0.0
How reproducible:
Create a project, select the sample (say nodejs-ex) and attempt to build it
Steps to Reproduce:
1.
2.
3.
Actual results: fails immediately, due to inability to start a pod
Expected results: pod starts, build runs, result image pushed
Additional info:
The project showing this behavior is https://console.stg.openshift.com/console/project/ldpjs/overview
Two things I see on this:
1. The build status doesn't seem to be properly conveyed. The build was failed but marked as completed.
2. A possible bug with the scheduler not taking node disk space into account - I pinged pmorie about that portion
The build status issue was already fixed here:
https://github.com/openshift/origin/pull/3936
this bug was opened for the scheduler issue, which is really an RFE more than a bug imho.