Title: Restarting Pods on Config Changes#

URL Source: https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html

Published Time: Thu, 30 Oct 2025 07:23:04 GMT

Markdown Content:
Housekeeping[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#housekeeping "Link to this heading")
---------------------------------------------------------------------------------------------------------------------------------------------

See the Getting Started documentation for installing and setting up:

*   microk8s

*   UCS Tools

    *   Accessing NGC

    *   Setting up repositories

Introduction[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#introduction "Link to this heading")
---------------------------------------------------------------------------------------------------------------------------------------------

In this tutorial, you will first demonstrate the problem mentioned in the previous tutorial [config changes not causing pods to restart](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/05_Partial_Redeployment_App.html#tutorial-5-configmap-issue) and show how to add dependencies on configmaps so that any changes in configmaps cause pods to restart.

This tutorial continues from [Partial Redeployment of App](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/05_Partial_Redeployment_App.html#tutorial-5).

Inspect current microservices and resources[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#inspect-current-microservices-and-resources "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 44m
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Running 0 2m16s

NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 44m

NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 44m
deployment.apps/http-server-http-server-deployment 1/1 1 1 44m

NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 44m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 1 1 1 44m

Also inspect the configmap contents where the `workingDir` parameter was added.

$ microk8s kubectl get cm -l "app.kubernetes.io/instance=server-client"
NAME DATA AGE
http-server-workload-cm 0 49m
curl-client-workload-cm 0 49m
http-server-configs-cm 2 49m
http-server-scripts-cm 1 49m
curl-client-scripts-cm 1 49m
curl-client-configs-cm 1 49m

$ microk8s kubectl get cm http-server-configs-cm -o yaml
apiVersion: v1
data:
 config.json: |
 {
 "workingDir": "/usr/local"
 }
 config.yaml: |
 SampleConfig:
 sampleValue: 0
kind: ConfigMap
metadata:
 annotations:
 meta.helm.sh/release-name: server-client
 meta.helm.sh/release-namespace: default
 creationTimestamp: "2023-03-14T06:05:18Z"
 labels:
 app.kubernetes.io/instance: server-client
 app.kubernetes.io/managed-by: Helm
 app.kubernetes.io/name: http-server
 app.kubernetes.io/version: 0.0.4
 generated_with: helm_builder
 hb_version: 0.7.1
 helm.sh/chart: http-server-0.0.4
 msb_version: 2.5.0
 name: http-server-configs-cm
 namespace: default
 resourceVersion: "954500"
 uid: 8bdae84f-84b0-411b-a7aa-337cab82bdc2

As you can see, `config.json` contains the current path of `/usr/local`.

Updating the parameters in the application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#updating-the-parameters-in-the-application "Link to this heading")
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Update the `app.yaml` file to the following:

specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application

dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.4

components:
- name: client
 type: ucf.svc.curl-client
- name: http-server
 type: ucf.svc.http-server
 parameters:
 workingDir: /var
 fileToCreate: someotherfile.txt

connections:
 client/http: http-server/http

The parameter `workingDir` has been modified and the app version has been updated.

Building the Application and Partial Redeployment[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#building-the-application-and-partial-redeployment "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Rebuild and redeploy the app using:

$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5

Inspecting the Microservices and Application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#inspecting-the-microservices-and-application "Link to this heading")
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 56m
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Running 0 15m

NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 57m

NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 56m
deployment.apps/http-server-http-server-deployment 1/1 1 1 56m

NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 56m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 1 1 1 56m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 53m

You can also get logs of the `curl-client` container:

$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
 % Total % Received % Xferd Average Speed Time Time Time Current
 Dload Upload Total Spent Left Speed
100 936 100 936 0 0 457k 0 --:--:-- --:--:-- --:--:-- 457k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="bin/">bin/</a></li>
<li><a href="cuda/">cuda@</a></li>
<li><a href="cuda-11/">cuda-11@</a></li>
<li><a href="cuda-11.4/">cuda-11.4/</a></li>
<li><a href="cuda-11.6/">cuda-11.6/</a></li>
<li><a href="etc/">etc/</a></li>
<li><a href="games/">games/</a></li>
<li><a href="include/">include/</a></li>
<li><a href="lib/">lib/</a></li>
<li><a href="man/">man@</a></li>
<li><a href="mpi/">mpi@</a></li>
<li><a href="nvm/">nvm/</a></li>
<li><a href="sbin/">sbin/</a></li>
<li><a href="share/">share/</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="src/">src/</a></li>
<li><a href="ucx/">ucx@</a></li>
</ul>
<hr>
</body>
</html>

Also get the contents of the `http-server` configmap:

$ microk8s kubectl get cm http-server-configs-cm -o yaml
apiVersion: v1
data:
 config.json: |
 {
 "workingDir": "/var"
 }
 config.yaml: |
 SampleConfig:
 sampleValue: 0
kind: ConfigMap
metadata:
 annotations:
 meta.helm.sh/release-name: server-client
 meta.helm.sh/release-namespace: default
 creationTimestamp: "2023-03-14T06:05:18Z"
 labels:
 app.kubernetes.io/instance: server-client
 app.kubernetes.io/managed-by: Helm
 app.kubernetes.io/name: http-server
 app.kubernetes.io/version: 0.0.4
 generated_with: helm_builder
 hb_version: 0.7.1
 helm.sh/chart: http-server-0.0.4
 msb_version: 2.5.0
 name: http-server-configs-cm
 namespace: default
 resourceVersion: "963732"
 uid: 8bdae84f-84b0-411b-a7aa-337cab82bdc2

You can see that the existing pod of `http-server` continues to run and return directory contents of the old path `/usr/local` even though the config reflects the updated path of `/var`.

Adding Dependency on configmaps[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#adding-dependency-on-configmaps "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Update the manifest of `http-server`:

type: msapplication
specVersion: 2.5.0
name: ucf.svc.http-server
chartName: http-server
description: http server
version: 0.0.5
tags: []
keywords: []
publish: false
ingress-endpoints:
 - name: http
 description: REST API endpoint
 protocol: TCP
 scheme: http
 mandatory: False
 data-flow: in-out

params:
 workingDir: "/localvol"
 #> type: string
 #> description: Working directory for listing contents
 #> flags: mandatory
 fileToCreate: somefile.txt
 #> type: string
 #> description: Name of the file to create in the working directory
 #> flags: mandatory
---
spec:
 - name: http-server-deployment
 type: ucf.k8s.app.deployment
 parameters:
 apptype: stateless

 - name: http-server-container
 type: ucf.k8s.container
 parameters:
 image:
 repository: nvcr.io/nvidia/pytorch
 tag: 22.04-py3
 command: [sh, -c]
 args: [
 "WORKING_DIR=$(cat /opt/configs/config.json | jq -r '.workingDir') && cd $WORKING_DIR && echo $PWD && touch $(FILE_TO_CREATE) && ls && python -m http.server 8080
 "]
 env:
 - name: POD_NAME
 valueFrom:
 fieldRef:
 fieldPath: metadata.name
 - name: FILE_TO_CREATE
 value: $params.fileToCreate
 ports:
 - containerPort: 8080
 name: http
 volumeMounts:
 - name: localvol
 mountPath: /localvol

 - name: svc
 type: ucf.k8s.service
 parameters:
 ports:
 - port: 8080
 protocol: TCP
 name: http

 - name: localvol
 type: ucf.k8s.volume
 parameters:
 persistentVolumeClaim:
 claimName: local-path-pvc

 - name: cm-dependencies
 type: ucf.appspec.restartPodOnConfigChanges
 parameters:
 # Add dependency on all configmaps detected in the microservice
 addAll: true

 # Add dependency on default configmaps for scripts, configs and workload configs
 # addDefault: true

 # Add dependency on individual configmaps using the configmap names
 # configMaps:
 # - http-server-configs-cm

You have two changes in the manifest:

*   A new component `cm-dependencies` of type `ucf.appspec.restartPodOnConfigChanges` has been added with `addAll` set to `true`

*   Microservice version has been updated

Update the `app.yaml` file to the following, only the `http-server` version has been updated in the `dependencies`:

specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application

dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.5

components:
- name: client
 type: ucf.svc.curl-client
- name: http-server
 type: ucf.svc.http-server
 parameters:
 workingDir: /var
 fileToCreate: someotherfile.txt

connections:
 client/http: http-server/http

Rebuild and Redeploy the Microservice and Application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#rebuild-and-redeploy-the-microservice-and-application "Link to this heading")
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Rebuild and redeploy the microservice and app using:

$ cd $HOME/ucf_tutorial/services
$ ucf_ms_builder_cli service build -d http-server/ -f
$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5

Inspect the Redeployed Application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#inspect-the-redeployed-application "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 78m
pod/http-server-http-server-deployment-d5764c767-vx6rp 1/1 Running 0 17s
pod/http-server-http-server-deployment-5cfdc4bfc9-24pkk 1/1 Terminating 0 37m

NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 78m

NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 78m
deployment.apps/http-server-http-server-deployment 1/1 1 1 78m

NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 78m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 75m
replicaset.apps/http-server-http-server-deployment-d5764c767 1 1 1 17s

You can see that the existing `http-server` pod has been terminated and a new pod launched. Checking the logs of the `curl-client` pod and the new `http-server` pod confirms that the new path `/var` took effect:

$ microk8s kubectl logs --tail -1 -l "app=http-server-http-server-deployment,microservice_version=0.0.5"
/var
backups
cache
lib
local
lock
log
mail
opt
run
someotherfile.txt
spool
tmp

$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
 % Total % Received % Xferd Average Speed Time Time Time Current
 Dload Upload Total Spent Left Speed
100 743 100 743 0 0 362k 0 --:--:-- --:--:-- --:--:-- 362k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="backups/">backups/</a></li>
<li><a href="cache/">cache/</a></li>
<li><a href="lib/">lib/</a></li>
<li><a href="local/">local/</a></li>
<li><a href="lock/">lock@</a></li>
<li><a href="log/">log/</a></li>
<li><a href="mail/">mail/</a></li>
<li><a href="opt/">opt/</a></li>
<li><a href="run/">run@</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="spool/">spool/</a></li>
<li><a href="tmp/">tmp/</a></li>
</ul>
<hr>
</body>
</html>

Update the Config Param and Rebuild and Redeploy the application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#update-the-config-param-and-rebuild-and-redeploy-the-application "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Update the `app.yaml` file to the following, only the `workingDir` parameter has been updated to `/tmp`:

specVersion: 2.5.0
version: 0.0.5
doc: README.md
name: server-client-app
description: Server Client Application

dependencies:
- ucf.svc.curl-client:0.0.2
- ucf.svc.http-server:0.0.5

components:
- name: client
 type: ucf.svc.curl-client
- name: http-server
 type: ucf.svc.http-server
 parameters:
 workingDir: /tmp
 fileToCreate: someotherfile.txt

connections:
 client/http: http-server/http

Rebuild and Redeploy the Application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#rebuild-and-redeploy-the-application "Link to this heading")
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Rebuild and redeploy the microservice and app using:

$ cd $HOME/ucf_tutorial/apps
$ ucf_app_builder_cli app build server-client-app/ -f
$ microk8s helm upgrade server-client server-client-app/server-client-app-0.0.5

Inspect the Redeployed Application[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#id1 "Link to this heading")
----------------------------------------------------------------------------------------------------------------------------------------------------------

$ microk8s kubectl get all
NAME READY STATUS RESTARTS AGE
pod/curl-client-curl-client-deployment-569849964-shcc2 1/1 Running 0 93m
pod/http-server-http-server-deployment-c88597df9-wr89f 1/1 Running 0 21s
pod/http-server-http-server-deployment-d5764c767-vx6rp 1/1 Terminating 0 15m

NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.152.183.1 <none> 443/TCP 49d
service/http-server-http-server-deployment-svc ClusterIP 10.152.183.245 <none> 8080/TCP 93m

NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/curl-client-curl-client-deployment 1/1 1 1 93m
deployment.apps/http-server-http-server-deployment 1/1 1 1 93m

NAME DESIRED CURRENT READY AGE
replicaset.apps/curl-client-curl-client-deployment-569849964 1 1 1 93m
replicaset.apps/http-server-http-server-deployment-7df656c66f 0 0 0 90m
replicaset.apps/http-server-http-server-deployment-5cfdc4bfc9 0 0 0 93m
replicaset.apps/http-server-http-server-deployment-c88597df9 1 1 1 21s
replicaset.apps/http-server-http-server-deployment-d5764c767 0 0 0 15m

You can see that the existing `http-server` pod has been terminated and a new pod launched. Checking the logs of the `curl-client` pod and the new `http-server` pod confirms that the new path `/tmp` took effect:

$ microk8s kubectl logs http-server-http-server-deployment-c88597df9-wr89f
/tmp
core-js-banners
someotherfile.txt
v8-compile-cache-0
yarn--1650154850728-0.1942747438338328
yarn--1650154852391-0.6050681633516224
yarn--1650154853881-0.6683114706726603
yarn--1650154920633-0.4394541481795069
yarn--1650154922130-0.5297801175307726

$ microk8s kubectl logs --tail -1 -l "app=curl-client-curl-client-deployment"
...
% Total % Received % Xferd Average Speed Time Time Time Current
 Dload Upload Total Spent Left Speed
100 989 100 989 0 0 482k 0 --:--:-- --:--:-- --:--:-- 482k
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>Directory listing for /</title>
</head>
<body>
<h1>Directory listing for /</h1>
<hr>
<ul>
<li><a href="core-js-banners">core-js-banners</a></li>
<li><a href="someotherfile.txt">someotherfile.txt</a></li>
<li><a href="v8-compile-cache-0/">v8-compile-cache-0/</a></li>
<li><a href="yarn--1650154850728-0.1942747438338328/">yarn--1650154850728-0.1942747438338328/</a></li>
<li><a href="yarn--1650154852391-0.6050681633516224/">yarn--1650154852391-0.6050681633516224/</a></li>
<li><a href="yarn--1650154853881-0.6683114706726603/">yarn--1650154853881-0.6683114706726603/</a></li>
<li><a href="yarn--1650154920633-0.4394541481795069/">yarn--1650154920633-0.4394541481795069/</a></li>
<li><a href="yarn--1650154922130-0.5297801175307726/">yarn--1650154922130-0.5297801175307726/</a></li>
</ul>
<hr>
</body>
</html>

This confirms that the change in configmap contents caused the pod to restart with the help of the `ucf.appspec.restartPodOnConfigChanges` component in the microservice manifest.

Stopping and Cleaning up Microservices and Applications[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#stopping-and-cleaning-up-microservices-and-applications "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Finally, to stop and clean up the application run:

$ microk8s helm3 uninstall server-client

Some Notes[#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#some-notes "Link to this heading")
-----------------------------------------------------------------------------------------------------------------------------------------

*   Only those configmaps that are part of the microservice helm chart can be added as dependencies.

Links/Buttons:
- [#](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/06_Restarting_Pods_Config_Changes.html#some-notes)
- [config changes not causing pods to restart](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/05_Partial_Redeployment_App.html#tutorial-5-configmap-issue)
- [Partial Redeployment of App](https://docs.nvidia.com/ucf/2.10.0/text/tutorials/05_Partial_Redeployment_App.html#tutorial-5)