Update node-feature-discovery deployment #108

e0ne · 2021-02-23T11:52:03Z

Use helm chart updated from NFD repo
network-operator specifies the latest build of the NFD
nfd-master '--instance' option is set to 'nvidia.networking' be default

Signed-off-by: Ivan Kolodiazhnyi <ikolodiazhny@nvidia.com>

moshe010 · 2021-02-23T19:03:06Z

/retest-nic_operator_helm

moshe010 · 2021-02-23T12:54:53Z

deployment/network-operator/charts/node-feature-discovery/Chart.yaml

@@ -1,8 +1,15 @@
 apiVersion: v2
-appVersion: v0.6.0
+appVersion: master


can't we put the image tag v0.6.0-233-g3e00bfb?

deployment/network-operator/charts/node-feature-discovery/templates/clusterrole.yaml

moshe010 · 2021-02-23T12:57:29Z

deployment/network-operator/charts/node-feature-discovery/templates/master.yaml

@@ -46,7 +46,7 @@ spec:
        - name: master
          securityContext:
            {{- toYaml .Values.master.securityContext | nindent 12 }}
-          image: "{{ .Values.image.repository }}:{{ .Chart.AppVersion }}"
+          image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default .Chart.AppVersion }}"


this to allow to change the tag? seem my comment above on the Chart.AppVersion which is master

Let's keep it as exists in NFD but hardcode version

Let's keep it as exists in NFD but hardcode version

+1
ideally no changes to the chart itself but only override values in network-operator values.yaml file

moshe010 · 2021-02-23T19:05:35Z

deployment/network-operator/charts/node-feature-discovery/templates/master.yaml

          resources:
            {{- toYaml .Values.master.resources | nindent 12 }}
+          args:
+            { { - if .Values.master.instance | empty | not } }
+            - "--instance={{ .Values.master.instance }}"


I don't think we want it to be configurable. you can hardcode the value nvidia.networking or something else

Let's hardcode it in values

moshe010 · 2021-02-23T19:05:54Z

deployment/network-operator/charts/node-feature-discovery/templates/nfd-worker-conf.yaml

@@ -0,0 +1,25 @@
+{{/*
+   Copyright 2020 NVIDIA


s/2020/2021

moshe010 · 2021-02-23T19:08:22Z

deployment/network-operator/values.yaml

@@ -24,6 +24,11 @@ sriovNetworkOperator:

 # Node Feature discovery chart related values
 node-feature-discovery:
+  image:
+    repository: gcr.io/k8s-staging-nfd/node-feature-discovery
+    tag: v0.6.0-233-g3e00bfb


can we do it deafult in the app vesion

moshe010 · 2021-02-23T19:14:53Z

deployment/network-operator/values.yaml

+    repository: gcr.io/k8s-staging-nfd/node-feature-discovery
+    tag: v0.6.0-233-g3e00bfb
+  master:
+    instance: "network-operator"


let remove this to be hard coded

One we'll start using chart from NFD helm repository, we won't be able to hardcode it without network-operator and/or admission controller changes, so I)'m not sure that it's a good idea to do it

abdallahyas · 2021-02-24T12:52:42Z

logs can be found here. it seems the pod was not labeled with pci-15b3.present or something similar

e0ne · 2021-02-24T14:44:44Z

/retest-nic_operator_helm

e0ne · 2021-02-24T15:28:04Z

deployment/network-operator/charts/node-feature-discovery/templates/worker.yaml

        args:
        - "--sleep-interval=60s"
        - "--server={{ include "node-feature-discovery.fullname" . }}-master:{{ .Values.master.service.port }}"
-        {{- with .Values.worker.options }}


It should be restored

adrianchiris

@e0ne i expect the nfd chart itself to be identical to what was merged to NFD with the exception of version override in node-feature-discovery/Chart.yaml

is that the case ?

adrianchiris · 2021-02-24T16:44:13Z

deployment/network-operator/charts/node-feature-discovery/templates/nfd-worker-conf.yaml

+  {{- include "node-feature-discovery.labels" . | nindent 4 }}
+data:
+  nfd-worker.conf: |
+  {{ .Values.worker.config | indent 4 }}


should we use the configmap and put our configs here and not under options WDYT ?

i.e override this value from the top level values.yaml in network-operator

Good catch. Let me try it

adrianchiris · 2021-02-24T16:47:58Z

deployment/network-operator/charts/node-feature-discovery/values.yaml

-  pullPolicy: IfNotPresent
+  repository: gcr.io/k8s-staging-nfd/node-feature-discovery
+  # This should be set to 'IfNotPresent' for released version
+  pullPolicy: Always


change in top level values.yaml to IfNotPresent ?

This patch also configures '--instance' option for nfd-master to be able to deploy several NFD instances in one cluster. Signed-off-by: Ivan Kolodiazhnyi <ikolodiazhny@nvidia.com>

e0ne · 2021-02-24T18:17:41Z

deployment/network-operator/values.yaml

-            - "0207"
-          deviceLabelFields:
-            - vendor
+          pci:


An indentation looks too big but it works as expected. Any suggestions are welcome

@e0ne please open an issue in github and in our RM for this.
I will merged as it has it doesn't effect functionality

adrianchiris · 2021-02-25T08:04:26Z

In regards to commit message:
This is actually the first time we use HELM chart from NFD, so commit message should reflect that.

Sync the latest Helm Chart for node-feature-discovery

26005d6

Signed-off-by: Ivan Kolodiazhnyi <ikolodiazhny@nvidia.com>

moshe010 requested changes Feb 23, 2021

View reviewed changes

e0ne force-pushed the nfd-instance-flag branch from 4d1e2b8 to bce0269 Compare February 24, 2021 09:52

e0ne changed the title ~~Update node-feature-discovery deployment~~ [WIP] Update node-feature-discovery deployment Feb 24, 2021

e0ne force-pushed the nfd-instance-flag branch 2 times, most recently from 2d7b5cc to 6bb436a Compare February 24, 2021 09:58

e0ne changed the title ~~[WIP] Update node-feature-discovery deployment~~ Update node-feature-discovery deployment Feb 24, 2021

e0ne commented Feb 24, 2021

View reviewed changes

e0ne force-pushed the nfd-instance-flag branch from 6bb436a to 665a083 Compare February 24, 2021 15:31

adrianchiris reviewed Feb 24, 2021

View reviewed changes

Use the latest NFD instance

744e369

This patch also configures '--instance' option for nfd-master to be able to deploy several NFD instances in one cluster. Signed-off-by: Ivan Kolodiazhnyi <ikolodiazhny@nvidia.com>

e0ne force-pushed the nfd-instance-flag branch from 665a083 to 744e369 Compare February 24, 2021 18:11

e0ne commented Feb 24, 2021

View reviewed changes

moshe010 approved these changes Feb 25, 2021

View reviewed changes

moshe010 merged commit 2dcf59f into Mellanox:master Feb 25, 2021

moshe010 mentioned this pull request Mar 1, 2021

Release v0.3.0 #124

Closed

26 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update node-feature-discovery deployment #108

Update node-feature-discovery deployment #108

e0ne commented Feb 23, 2021 •

edited

Loading

moshe010 commented Feb 23, 2021

moshe010 Feb 23, 2021

moshe010 Feb 23, 2021

e0ne Feb 24, 2021

adrianchiris Feb 24, 2021

moshe010 Feb 23, 2021

e0ne Feb 24, 2021

moshe010 Feb 23, 2021

moshe010 Feb 23, 2021

moshe010 Feb 23, 2021

e0ne Feb 24, 2021

abdallahyas commented Feb 24, 2021

e0ne commented Feb 24, 2021

e0ne Feb 24, 2021

adrianchiris left a comment •

edited

Loading

adrianchiris Feb 24, 2021 •

edited

Loading

e0ne Feb 24, 2021

adrianchiris Feb 24, 2021

e0ne Feb 24, 2021 •

edited

Loading

moshe010 Feb 25, 2021

adrianchiris commented Feb 25, 2021

Update node-feature-discovery deployment #108

Update node-feature-discovery deployment #108

Conversation

e0ne commented Feb 23, 2021 • edited Loading

moshe010 commented Feb 23, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

abdallahyas commented Feb 24, 2021

e0ne commented Feb 24, 2021

Choose a reason for hiding this comment

adrianchiris left a comment • edited Loading

Choose a reason for hiding this comment

adrianchiris Feb 24, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

e0ne Feb 24, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

adrianchiris commented Feb 25, 2021

e0ne commented Feb 23, 2021 •

edited

Loading

adrianchiris left a comment •

edited

Loading

adrianchiris Feb 24, 2021 •

edited

Loading

e0ne Feb 24, 2021 •

edited

Loading