Step 6: Install OpenLM Platform

Install the openlm-platform Helm release, then follow the pods until every one is running. Load the variables from openlm-install.env, the file you created from the installation worksheet:

source ./openlm-install.env

6.1 Install the release

Install the release with Helm:

helm upgrade --install openlm-platform openlm/openlm-platform \
  --version "$OPENLM_VERSION" -n "$NS" -f openlm-values.yaml --timeout 60m

OpenShift

On OpenShift, add your second values file: -f openlm-values.yaml -f openlm-values-openshift.yaml. After the command finishes, follow the Route steps.

The command returns when the chart's topic setup job finishes. That job creates or checks each of OpenLM Platform's Kafka topics, one after the other, and takes 10 to 30 minutes, so the command looks idle for a while. This is expected. Don't press Ctrl+C. The job runs again on every upgrade, even when no topic changes, so always keep --timeout 60m and plan the time into upgrade windows.

Pre-release versions

A version such as 1.0.0-rc.4 is a pre-release. helm search repo openlm lists it only when you add --devel, but installing it with --version works as shown.

6.2 Follow the progress

To follow the progress, watch the pods in a second terminal:

kubectl -n openlm get pods -w

The pods go through these stages, in order:

1. Pods appear and start (first few minutes). Many go through ContainerCreating while images download.
2. openlm-identity takes longest to become ready, up to 15 minutes on the first start, because it creates its database tables. Other services might restart a few times while they wait for it. This is expected.
3. openlm-usersandgroupsservice also has a long first start. Let it run.
4. The topic setup job (kafkainitpod) runs and then shows Completed.

6.3 Check that the installation finished

The installation has finished when the following command prints nothing, and every Running pod shows all containers ready (1/1):

kubectl -n "$NS" get pods --no-headers | grep -vE "Running|Completed"

On a cluster that pulls every image for the first time, plan 20 to 40 minutes.

Fix pods that don't become ready

If a pod stays in an error or waiting status, find its status in the following table:

Pod status	Most likely cause	Where to look
Pending for more than 5 minutes	No node fits: not enough capacity, or node placement is on but nodes aren't labeled	kubectl -n $NS describe pod <pod>, last lines. Node placement
ImagePullBackOff	Nodes can't reach public.ecr.aws, or your mirror lacks the image	Networking: registry
CreateContainerConfigError	A secret is missing	Run step 3 again
CrashLoopBackOff on many pods	A data service is unreachable or a credential is wrong	kubectl -n $NS logs <pod> --previous
CrashLoopBackOff with IdentityUnavailable or SSL in the log	Pods can't reach or don't trust https://$DOMAIN/identity	Sign-in and identity
helm stops with timed out waiting for the condition	The topic setup job didn't finish in time, usually because Kafka is slow or unreachable	kubectl -n $NS logs job/kafka-init-pod, then run the same helm upgrade --install again with --timeout 60m. It continues, and the topics already created stay.

For more problems and fixes, see Troubleshooting.

Check

Check the release status:

helm -n "$NS" status openlm-platform | grep STATUS

The output must say STATUS: deployed, and the pod check in 6.3 must print nothing.