Fail to download pod containers when proxy is in place about dcos-kubernetes-quickstart HOT 14 CLOSED

@ezreal92170 I just installed beta-kubernetes package v0.2.2-1.7.7-beta without any problem. However it is possible kube-dns isn't ready whenever we try to deploy it in the first iteration getting this error kube-dns check failed and is not ready. The system creates a new task mandatory-addons-0-kube-dns that retries to install kube-dns again. Can you confirm you don't see any mandatory-addons-0-kube-dns task with Finished status ? If so, could you run $ kubectl get pods -n kube-system to give us more details ?

from dcos-kubernetes-quickstart.

Yes I don't see any mandatory-addons-0-kube-dns task with Finished status..

kubectl get pods -n kube-system
NAME READY STATUS RESTARTS AGE
kube-dns-2102953216-dvw4g 0/3 ContainerCreating 0
`
And I have the following error log on kube node :

E1015 19:27:04.550963 53 remote_runtime.go:91] RunPodSandbox from runtime service failed: rpc error: code = 2 desc = failed pulling image "gcr.io/google_containers/pause-amd64:3.0": Error response from daemon: {"message":"Get https://gcr.io/v1/_ping: dial tcp 173.194.208.82:443: i/o timeout"}
E1015 19:27:04.551042 53 kuberuntime_sandbox.go:54] CreatePodSandbox for pod "kube-dns-2102953216-dvw4g_kube-system(66e22e9c-b1cd-11e7-9aac-801844e90be0)" failed: rpc error: code = 2 desc = failed pulling image "gcr.io/google_containers/pause-amd64:3.0": Error response from daemon: {"message":"Get https://gcr.io/v1/_ping: dial tcp 173.194.208.82:443: i/o timeout"}

I need to use http proxy to access Internet. My issue on this subject was closed but maybe it's still an issue.

from dcos-kubernetes-quickstart.

@ezreal92170 yeah, it seems it cannot pull the image. That is why it doesn't work.

from dcos-kubernetes-quickstart.

@ezreal92170 are you behind a proxy?

@spahl aren't we supposed to be air-gap friendly?

from dcos-kubernetes-quickstart.

If you are behind a proxy or do not have direct internet access from the nodes then you need to build a local universe containing the beta-kubernetes package and use that.

More here: https://dcos.io/docs/1.10/administering-clusters/deploying-a-local-dcos-universe/#-a-name-build-a-installing-your-own-set-of-universe-packages

from dcos-kubernetes-quickstart.

I am behind a proxy but I don't understand why building a local universe could help.
I have no problem to create the framework.

from dcos-kubernetes-quickstart.

The problem seems to be that the framework scheduler can't download the artifacts it needs, namely the official Kubernetes components, because it doesn't know about your proxy settings. This is why we have what @spahl linked above.

from dcos-kubernetes-quickstart.

Should be fixed in 0.3.0-1.7.10-beta release, for as long as you have DC/OS configured to use the HTTP proxy.

from dcos-kubernetes-quickstart.

I'm having the same issue with 0.4.0-1.9.0-beta. All artifacts seems to be downloaded in the work dir.

2018/01/11 10:58:59 Resolve disabled via -resolve=false: Skipping host resolution
2018/01/11 10:58:59 Template handling disabled via -template=false: Skipping any config templates
2018/01/11 10:58:59 No $MESOS_SANDBOX/.ssl directory found. Cannot install certificate. Error: stat /var/lib/mesos/slave/slaves/59336e89-bac9-44c0-ba44-78a66392b7b1-S8/frameworks/56308059-3e74-48f4-86a9-67f9bd519f44-0001/executors/dev_k8.69908f7c-f6be-11e7-8ff9-fe0d65b1eb4a/runs/82acd8eb-26f1-47c8-9dce-b72d3cfbb954/.ssl: no such file or directory
2018/01/11 10:58:59 Local IP --> 192.168.5.91
2018/01/11 10:58:59 SDK Bootstrap successful.
Exception in thread "main" com.mesosphere.sdk.kubernetes.scheduler.tls.exceptions.UnrecoverableTLSException: com.mesosphere.sdk.state.StateStoreException: Key cannot contain '/' (reason: LOGIC_ERROR)
	at com.mesosphere.sdk.kubernetes.scheduler.tls.StateStoreTLSStore.loadServiceAccountKeyPair(StateStoreTLSStore.java:69)
	at com.mesosphere.sdk.kubernetes.scheduler.tls.OpenTLSProvisioner.initializeServiceAccountKeyPair(OpenTLSProvisioner.java:64)
	at com.mesosphere.sdk.kubernetes.scheduler.tls.TLSProvisioner.initialize(TLSProvisioner.java:63)
	at com.mesosphere.sdk.kubernetes.scheduler.Main.main(Main.java:61)
Caused by: com.mesosphere.sdk.state.StateStoreException: Key cannot contain '/' (reason: LOGIC_ERROR)
	at com.mesosphere.sdk.state.StateStore.validateKey(StateStore.java:582)
	at com.mesosphere.sdk.state.StateStore.fetchProperty(StateStore.java:387)
	at com.mesosphere.sdk.kubernetes.scheduler.tls.StateStoreTLSStore.loadServiceAccountKeyPair(StateStoreTLSStore.java:60)
	... 3 more
I0111 10:59:21.747092    11 executor.cpp:925] Command exited with status 1 (pid: 15)
I0111 10:59:22.748448    14 process.cpp:1072] Failed to accept socket: future discarded

from dcos-kubernetes-quickstart.

@ezreal92170 https://docs.mesosphere.com/services/beta-kubernetes/0.4.0-1.9.0-beta/advanced-install/#when-you-have-a-proxy

from dcos-kubernetes-quickstart.

@deric that's a different issue and i'll open a new one with your feedback.

from dcos-kubernetes-quickstart.

@pires was new issue created for @deric 's issue?
thanks!

from dcos-kubernetes-quickstart.

@cap1981 #52

We are about to release a new beta that includes a lot of bug-fixes. Stay tuned.

from dcos-kubernetes-quickstart.

Thank you! i am dealing with similar issues with mandatory-addon.

from dcos-kubernetes-quickstart.