trouble running kube-up.sh

[Jonathan Wills]

I just upgraded to kubernetes v1.0.1 and I'm having trouble running kube-up.sh.  The loop at the end never terminates, and when I ssh into the master to see what's up this is what I fine

Broken (or in progress) GCE Kubernetes node setup! Suggested first step: tail /var/log/startupscript.logrunningwild@kubernetes-master:~$ tail /var/log/startupscript.logAug 1 16:48:03 kubernetes-master startupscript: File "/usr/lib/python2.7/dist-packages/yaml/scanner.py", line 149, in need_more_tokensAug 1 16:48:03 kubernetes-master startupscript: self.stale_possible_simple_keys()Aug 1 16:48:03 kubernetes-master startupscript: File "/usr/lib/python2.7/dist-packages/yaml/scanner.py", line 289, in stale_possible_simple_keysAug 1 16:48:03 kubernetes-master startupscript: "could not found expected ':'", self.get_mark())Aug 1 16:48:03 kubernetes-master startupscript: yaml.scanner.ScannerError: while scanning a simple keyAug 1 16:48:03 kubernetes-master startupscript: in "<stdin>", line 28, column 1Aug 1 16:48:03 kubernetes-master startupscript: could not found expected ':'Aug 1 16:48:03 kubernetes-master startupscript: in "<stdin>", line 29, column 1Aug 1 16:48:03 kubernetes-master startupscript: /tmp/tmp.fkEKCb3yOF: line 596: KUBERNETES_MASTER: unbound variableAug 1 16:48:03 kubernetes-master startupscript: Finished running startup script /var/run/google.startup.script

I tried this at v1.0.0 and v1.0.1.  Have I done something wrong?  The only changes I made to the configs before running kube-up.sh was to change the number of minions and the region.

[Brendan Burns]

Jonathan,

Does it work for you if you don't change the default config?  You can also set environment variables for number of minions and zone rather than edit files, perhaps try that?

If you can send me the details of what you are trying, I can try to reproduce and see if I can figure out what is happening.

Thanks

--brendan

--
You received this message because you are subscribed to the Google Groups "Containers at Google" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-contain...@googlegroups.com.
To post to this group, send email to google-c...@googlegroups.com.
Visit this group at http://groups.google.com/group/google-containers.
For more options, visit https://groups.google.com/d/optout.

[Jonathan Wills]

I have a micro instance on GCE, I logged in, downloaded a new kubernetes, and ran kube-up.sh.  I've used older versions successfully from this instance.  I've restarted the machine, and I tried a fresh kubernetes without changing the configs, same error in all cases. 

I just realized that the error message that I posted did not format at all.  I think this is the relevant part:

Aug 1 16:48:03 kubernetes-master startupscript: /tmp/tmp.fkEKCb3yOF: line 596: KUBERNETES_MASTER: unbound variable

versions, if they matter:

$ gcloud version
Google Cloud SDK 0.9.71


alpha 2015.07.24
bq 2.0.18
bq-nix 2.0.18
core 2015.07.24
core-nix 2015.06.02
gcloud 2015.07.24
gsutil 4.13
gsutil-nix 4.12
kubectl
kubectl-linux-x86_64 1.0.1
preview 2015.07.24

[Jonathan Wills]

Ah, I'd forgotten that I'd had to set some environment variables to try out rocket.  I had this in my env:

KUBE_GCE_MINION_IMAGE=coreos-stable-717-3-0-v20150710
KUBE_GCE_MINION_PROJECT=coreos-cloud
KUBE_CONTAINER_RUNTIME=rkt
KUBE_OS_DISTRIBUTION=coreos

So I guess now my question is how to get rocket working :)

I'm following the directions here.  I noticed that I didn't have the appropriate version of systemd installed, so now I'm trying to install a higher version and then I'll try again.  I'm having trouble getting a high version though, but I'll post again when I've tried bringing up a rkt cluster again.

[Jonathan Wills]

Just to be clear, when I removed those variables from my env I was able to bring up a cluster without a problem.

On Saturday, August 1, 2015 at 1:36:40 PM UTC-4, Jonathan Wills wrote:

[Jonathan Wills]

Ok, in order to get systemd up to the version you guys require I actually had to change to a VM running ubuntu 15.04, but that wasn't a big deal.  After reinstalling everything and setting those environment variables again I still get exactly the same error.

Any ideas?

[Jonathan Wills]

I found the problem.  The kube-env yaml file that the master uses on startup has a typo in it (it uses a '=' rather than a ':' for one of the values).

I made PR 12163 to fix this.