add routes, persistant storage, environment

[holgerkoch]

• add shared persistant storage to all nodes, mounted /data
• add environment (NSS_WRAPPER_PASSWD, NSS_WRAPPER_GROUP, LD_PRELOAD) to all nodes
• create (generated) routes

[crobby]

Thanks for the PR. I, and many others that might be interested in reviewing this are currently at KubeCon Europe, so we may be slightly delayed in our reviews.

[holgerkoch]

Hi,

no Problem! I'm already there for the OpenShift Commons Gathering.

Best regards

Holger

Von meinem iPhone gesendet

Am 28.03.2017 um 11:59 schrieb Chad Roberts notifications@github.com:

Thanks for the PR. I, and many others that might be interested in reviewing this are currently at KubeCon Europe, so we may be slightly delayed in our reviews.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[elmiko]

thanks for the PR @holgerkoch !

[mattf]

@holgerkoch i hope the origin commons gathering was helpful.

the template you're editing here isn't the template we recommend using. instead we recommend the oshinko tooling (webui or cli), see http://radanalytics.io/get-started, which will generate a template for you based on some input, even input of a configmap for cluster config.

re persistent volumes, as folks mentioned, we're currently expecting most apps to use "app level" storage, e.g. app connects to kafka:// or hdfs:// or s3://, instead of persistent volumes, e.g. file:// via pvc/pv from existing nfs. that said, would you look at the oshinko tooling (https://github.com/radanalyticsio/oshinko-cli) and make a proposal on how you would like to deploy a cluster w/ persistent volumes attached?

re nss_wrapper, is this because the developers connect to a cluster and get the dreaded "failure to login" exception because the ephemeral uid isn't in /etc/passwd? if so, we recommend the developers use oc run instead of oc rsh/exec. take a look at http://radanalytics.io/faq for an example of oc run and explanation of how to use oc rsh/exec.

[holgerkoch]

@mattf thank you very much for your information. I don't know the radanalytics project. It sounds really interesting and i like to take a deeper look. I will talk with our developers, why they use persistent storage. Maybe there is no reason and they also never heard from radanalytics. You are right with the nss_wrapper point. They login with "oc rsh", set the environment and start a spark shell. Maybe this is a wrong way. I also talk about this with the developers. Apache Spark is a very new business for us.

I found a list with some (maybe all) default icons. https://rawgit.com/openshift/origin-web-console/master/app/styles/fonts/openshift-logos-icon/demo.html

Best regards

Holger

[elmiko]

would be cool if we could get a spark icon in there =)

maybe just the apache logo would work for now

[mattf]

@holgerkoch re "oc rsh" - "oc run" it is strongly recommended instead "oc rsh". it will give them an independent pod, which avoids perturbing the spark cluster, as well as more closely mimics how their eventual application will run. if the developers are prototyping, they might also try running a notebook like jupyter in a pod and connecting it to a cluster. you can find a simple example in http://radanalytics.io/applications/s3-source-example

[tmckayus]

I'm going to close this PR since it's been open for some time and has branch conflicts. I agree with comments above that in general additional features should be added through radanalyticsio tooling and this image should/repo should be used for a basic spark cluster, potentially with optional features (like metrics) that can be controlled with env vars. Thanks!