[Starlingx-discuss] [Containers] node-feature-discovery

[Perez Carranza, Jose]

Hi Cris 

Thanks for your info. I was able to install node-feature-discovery, following the requirements described on storyboard [1], I'm checking that I should be able to deploy a "demonset" for master and worker nodes, Can I use the template defined on github [2]? Basically I' following the instruction provided there there  and running below: 

 
`sed -E s',^(\s*)image:.+$,\1image: quay.io/kubernetes_incubator/node-feature-discovery:v0.3.0,' nfd-master.yaml.template > nfd-master.yaml && kubectl create -f nfd-master.yaml` 

But I'm geting below error:

`$ kubectl create -f nfd-master.yaml
Unable to connect to the server: net/http: TLS handshake timeout` 

and now I cannot execute any other kubectl command

1. https://storyboard.openstack.org/#!/story/2005193
2. https://github.com/kubernetes-sigs/node-feature-discovery/

Thanks 
Jose 
________________________________________
From: Chris Friesen [chris.friesen at windriver.com]
Sent: Monday, June 03, 2019 2:35 PM
To: Perez Carranza, Jose; 'starlingx-discuss at lists.starlingx.io'
Subject: Re: [Containers] node-feature-discovery

Hi José,

The docs are still in progress.  Here's what I sent to the docs folks:

The intent of this story was to deploy the functionality implemented by
https://github.com/kubernetes-sigs/node-feature-discovery/tree/v0.3.0 as
an optional component in StarlingX.

Basically it detects hardware features available on each node in a
Kubernetes cluster, and advertises those features using kubernetes node
labels.  You can then use the regular Kubernetes label-based
functionality to specify the nodes with features that are of interest.
(nodeSelector, "kubectl get node -l <criteria>", etc.)

As of the beginning of May StarlingX should have a new file
/opt/extracharts/node-feature-discovery-0.3.0.tgz which is a helm chart
that provides the above functionality.  Our version also allows for
customization of various parameters via helm chart overrides.  The
configurable options and their defaults are as follows:

# namespace to use for chart resources.  Must be specified.
namespace: default

# label for the daemonset to find its pods
app_label: node-feature-discovery

# docker image to use for the pods
image: quay.io/kubernetes_incubator/node-feature-discovery:v0.3.0

# interval (in secs) to scan the node features
scan_interval: 60

# key/value pair to match against node labels to select which nodes
# should run the node feature discovery.  Defaults to all nodes.
node_selector_key:
node_selector_value:


In the simple case where we want to run using all the default values,
after initial install and configuration the helm chart can be installed
by running "helm upgrade -i node-feature-discovery
/opt/extracharts/node-feature-discovery-0.3.0.tgz".  This should result
in the creation of one pod per node which runs once per minute to update
the node features.

Thanks,
Chris


On 6/3/2019 9:11 AM, Perez Carranza, Jose wrote:
> Hi Chris
>
> Just to follow up on below question, do you have any info to share about node-feature-discovery feature?
>
> Regards,
> José
>
>
>> -----Original Message-----
>> From: Perez Carranza, Jose
>> Sent: Wednesday, May 29, 2019 11:34 AM
>> To: Chris Friesen <chris.friesen at windriver.com>; starlingx-
>> discuss at lists.starlingx.io
>> Cc: Miller, Frank <Frank.Miller at windriver.com>
>> Subject: [Containers] node-feature-discovery
>>
>> Hi Chris
>>
>> I'm checking storyboard for node-feature-discovery  [1] to design test scenarios
>> about it, for me is not clear yet how to enable that feature on my deployment,
>> are you able to explain more on how to do it? I also see that documentation
>> was provided to docs team, are you able to point me out to that
>> documentation so I can have more details of this feature implementation.
>>
>> 1. https://storyboard.openstack.org/#!/story/2005193
>>
>> Regards,
>> José
>>
>