Re: [atomic] How To Create a Node for Atomic Registry to Deploy to

[Seth Jennings <sjenning redhat com>]

I am also seeing this.

The basic issue is that there are no origin nodes, only an origin
master.  This leads to the origin underlay failing to schedule the
docker-registry deployer:

atomic-registry-master[2579]: I0324 19:32:22.524224       1
scheduler.go:97] Failed to schedule pod:
default/docker-registry-1-deploy

Not sure how this ever worked, unless, at some point in the past there
really was an all-in-one origin service with both master and node.
But the atomic-registry-master.service now is just the master (grep
for "start master" below):

# cat /etc/systemd/system/atomic-registry-master.service | grep ExecStart
ExecStart=/usr/bin/docker run --rm --privileged --name
atomic-registry-master -p ${MASTERPORT}:${MASTERPORT}
--env-file=/etc/sysconfig/atomic-registry-master -v
/var/lib/atomic-registry/etcd:/var/lib/atomic-registry/etcd -v
/etc/atomic-registry/:/etc/atomic-registry/ -v
/etc/pki/ca-trust:/etc/pki/ca-trust:ro ${MASTERIMAGE}:${MASTERTAG}
start master --config=/etc/atomic-registry/master/master-config.yaml
$OPTIONS

So afaict, this installation method is totally broken right now.

Seth

On Fri, Jan 27, 2017 at 1:03 PM, Eric <naisanza gmail com> wrote:
> I'm using Fedora's Atomic distribution, and following the Registry setup
> guide here:
>
> http://docs.projectatomic.io/registry/latest/registry_quickstart/administrators/index.html#registry-quickstart-administrators-index
>
> Following it step by step. The Registry services work, but when creating a
> new Deployment it hangs because it's unable to find any "available nodes" to
> deploy on.
>
> I would like to have it working on Fedora Atomic instead of OpenShift. Is
> there something else that I'm missing that I need to do before setting up
> Registry? kubeamd init?
>
>
> On January 27, 2017 1:53:44 PM EST, Aaron Weitekamp <aweiteka redhat com>
> wrote:
>>
>> The atomic registry installer[1] deploys an all-in-one registy-master-node
>> configuration. It's not intended to support arbitrary pod deployment. The
>> installer should deploy the registry and console pods on the node.
>>
>> However, the best supported deployment today is the stand-alone registry
>> method on OpenShift[2]. Slightly more complex but more reliable, better
>> tested, etc.
>>
>> [1]
>> http://docs.projectatomic.io/registry/latest/registry_quickstart/administrators/index.html
>> [2]
>> https://docs.openshift.org/latest/install_config/install/stand_alone_registry.html
>>
>> On Thu, Jan 26, 2017 at 2:54 PM, Eric <naisanza gmail com> wrote:
>>>
>>> Following the Quick Start Guide doesn't tell you how to create a Node.
>>> I'm unable to deploy any images since it requires a Node to deploy on
>>>
>>> When creating a new Deployment I get the error in the event messages:
>>>
>>> "no nodes available to schedule pods"
>>>
>>>
>>> I've also opened up a question here:
>>> http://ask.projectatomic.io/en/question/3878/atomic-registry-how-to-create-a-node/
>>
>>
>