deployment

[phette23]

Update 2024-10-04: Mark is working on deployment and leaning towards not using helm. Below are my notes on getting the Invenio helm chart to work.

---

See helm-invenio but helm repo add doesn't work yet. We'll write our own values.yml file. Deployments will be like cat values.yaml | envsubst | helm install --namespace $NS invenio /path/to/helm-invenio/charts/invenio -f - and we populate the env with our secrets.

Local issues with minikube:

• [x] opensearch pods readiness probe failing (try disabling them) — problem went away on its own!
• [x] image architecture mismatch for invenio pods (build my own)
  • [x] add image pull secret
  • [x] script to tag and push image to Artifact Registry (see below)
• [ ] exec error on invenio containers (probably from building on MacOS? try to explicitly specify arch)

General outline:

• [x] create deploy dir for main script and values file
• [x] ci.sh script for everything like cca.edu does
• [x] scripts builds and pushes to AR
• [ ] script does a helm install/upgrade
• [x] .gitlab-ci.yml with tasks for builds and helm deploys

[phette23]

I'm unable to get the helm chart running locally on minikube. The base almalinux image from CERN is available for arm64 and amd64 architectures docker manifest inspect --verbose registry.cern.ch/inveniosoftware/almalinux:latest. Minikube runs as aarch64 minikube ssh 'uname -m' but for some reason even the arm64 image does not run:

¿ minikube ssh
docker@minikube:~$ docker pull registry.cern.ch/inveniosoftware/almalinux:latest
docker@minikube:~$ docker inspect registry.cern.ch/inveniosoftware/almalinux:latest | grep Architecture
        "Architecture": "arm64",
docker@minikube:~$ docker run -it registry.cern.ch/inveniosoftware/almalinux:latest sh
exec /usr/bin/sh: exec format error

I've tried using a qemu driver for minikube instead but the problem persists.