SWM Installation Fails after Mar 23, 2025 commit

Hi @jkgesswein, after latest commits, the SWM fails to install:

SWM installation logs

2025-03-27 09:27:56 (9.54 MB/s) - '/usr/bin/yq' saved [8949760/8949760]

make[1]: Entering directory '/qos-scheduler'
mkdir -p tmp/
sed -ne "1,2p" <helm/qos-scheduler/Chart.yaml >tmp/header-yaml.txt
controller-gen paths="./scheduler/..." output:rbac:artifacts:config=helm/qos-scheduler/charts/controller/templates/ r                                                                                                                        bac:headerFile="tmp/header-yaml.txt",roleName=manager-role
/bin/sh: 1: controller-gen: not found
make[1]: *** [Makefile:786: helm/qos-scheduler/charts/controller/templates/role.yaml] Error 127
make[1]: Leaving directory '/qos-scheduler'
Error: INSTALLATION FAILED: repo tmp not found

It seems that is misses controller-gen, which I installed manually. This way make chart works, but helm install qostest failed, like bellow:

Logs after manually installing controller-gen

root@zorro-solutions:/qos-scheduler# make chart
mkdir -p tmp/
sed -ne "1,2p" <helm/qos-scheduler/Chart.yaml >tmp/header-yaml.txt
controller-gen paths="./scheduler/..." output:rbac:artifacts:config=helm/qos-scheduler/charts/controller/templates/ rbac:headerFile="tmp/header-yaml.txt",roleName=manager-role
mkdir -p deliv/helm/
mkdir -p tmp/helm
rsync -qru --delete helm/qos-scheduler/ tmp/helm
for f in tmp/helm/charts/qos-crds/crds/*.yaml ; do\
    yq --inplace 'del(.metadata.annotations)' "${f}" ;\
done
for src in helm/qos-scheduler/values.yaml helm/qos-scheduler/charts/*/values.yaml ; do\
    sed -Ee "/^ +pullPolicy:/s/:.+/: Always/" \
        -e "1,/^# External/{/^ +tag:/s/:.+/: 2.0.3/;}" \
    ${src} >tmp/helm${src#"helm/qos-scheduler"} ; \
done
helm dependency update tmp/helm
Getting updates for unmanaged Helm repositories...
...Successfully got an update from the "https://charts.jetstack.io" chart repository
...Successfully got an update from the "https://charts.bitnami.com/bitnami" chart repository
Saving 15 charts
Downloading multus-cni from repo https://charts.bitnami.com/bitnami
Downloading cert-manager from repo https://charts.jetstack.io
Deleting outdated charts
rm `find tmp/helm -type f -name "*-0.0.0.tgz"`
helm package --app-version=2.0.3 --version=2.0.3 \
     -d deliv/helm/ tmp/helm
Successfully packaged chart and saved it to: deliv/helm/qos-scheduler-2.0.3.tgz
helm repo index deliv/helm/ --url /projects//packages/helm/api/stable/charts
root@zorro-solutions:/qos-scheduler# helm install qostest --namespace=he-codeco-swm --create-namespace --set multus-cni.enabled=false tmp/helm
Error: INSTALLATION FAILED: template: qos-scheduler/charts/solver/templates/service.yaml:13:73: executing "qos-scheduler/charts/solver/templates/service.yaml" at <"solver">: nil pointer evaluating interface {}.uid

assigned to @jkgesswein

mentioned in issue #9 (closed)

Hi @jkgesswein ,

should I add the following lines in the post_deploy.sh to install controller-gen, or is it something that should be dealt with through by SWM? This is currently missing, causing the SWM installation to fail.

go install sigs.k8s.io/controller-tools/cmd/controller-gen@v0.15.0
export PATH=$(go env GOPATH)/bin:$PATH

I am working on a better solution for Scheduler’s installation. It does no longer require building Scheduler’s chart locally but “loads” it using a container. I am testing this already. With the new solution, it is not necessary to install controller-gen.

I think there is another problem with Scheduler’s installation masked by the issue with controller-gen that will be resolved by the fix. So, just stay tuned…

@karagk @karamolegkos
just a confirmation that this issue matches with what i see on my machine.

deploying acm after manually adding the controller-gen gives:
Error: INSTALLATION FAILED: template: qos-scheduler/charts/solver/templates/service.yaml:13:73: executing "qos-scheduler/charts/solver/templates/service.yaml" at <"solver">: nil pointer evaluating interface {}.uid

( no he-codeco-swm pods created )

it looks like the same to me

Hi @jkgesswein ,

I am trying to deploy SWM after your latest update. Based on the guide here, I am running the following:

cd qos-scheduler
REGISTRY_PREFIX=hecodeco/swm-
eval "$(docker run --rm "${REGISTRY_PREFIX}chart-dh:$(cat VERSION)"  install --release=qostest)"

but I am getting

Unknown command: “chart”

Can you help me properly deploy the new version?

added High Priority label

another note on failed install (might be an acm issue instead) .

When i deployed SWM via the acm/scripts/post_deploy.sh it gives me:

helm install qostest --namespace=he-codeco-swm --create-namespace --set multus-cni.enabled=false tmp/helm
Error: INSTALLATION FAILED: repo tmp not found

i could "fix" this by running

helm install qostest --namespace=he-codeco-swm --create-namespace tmp/helm-gl

But i dont know what "gl" is and if that creates a correct environment.

Hi all,

@jkgesswein provided me the correct commands to deploy SWM:

docker run hecodeco/swm-chart-dh:2.0.4 >chart.tgz
docker run --rm hecodeco/swm-chart-dh:2.0.4 help # optional, just shows available commands
docker run --rm hecodeco/swm-chart-dh:2.0.4 install # this outputs the following installation command, but is broken

# This is what basically does the installation, but is to be refined:
( fn=$(mktemp "/tmp/chart-XXXXXX") && docker run --rm "hecodeco/swm-chart-dh:2.0.4" >"${fn}" && helm install "qos" "${fn}" --namespace="he-codeco-swm" --create-namespace && rm -f "${fn}" )

This successfully deployed all pods, I will do further testing tomorrow regarding codecoApps and node recommendations.

@tymon and @karamolegkos have a look!

closed